Technology Assessment: Using Biometrics for Border Security
(14-NOV-02, GAO-03-174).
As directed in the Fiscal Year 2002 Legislative Branch
Appropriations Conference Report and subsequent support letters
from interested members of Congress, GAO conducted a pilot
program in technology assessment that examined the use of
biometric technologies for border control. Biometric technologies
are available today and are being used for a variety of
applications such as access control and criminal identification
and surveillance. GAO considered a number of leading and emerging
biometric technologies that could potentially be used for
securing the nations borders. The seven leading biometric
technologies include facial recognition, fingerprint recognition,
hand geometry, iris recognition, retina recognition, signature
recognition, and speaker recognition. To evaluate the
effectiveness of biometrics in border control, it is important to
recognize that the use of biometric technology would be but one
component of the decision to support systems that determine who
is allowed to enter the United States and who is not. Biometric
technology can play a role in associating a person with travel
documents such as visas and passports. When used at a border
inspection, the biometric comparison can be used to help decide
whether to admit a traveler into the United States. Before any
decision is made to implement biometrics in a border control
system, the benefits of the system must be weighed against its
costs. The purpose of any biometrics initiative is to prevent the
entry of travelers who are inadmissible to the United States. The
costs of biometric border control system will not be trivial.
Important policy implications must be addressed in trade-offs
between increasing security and the impact on areas such as
privacy, economy, traveler convenience, and international
relations. Civil liberties groups and privacy experts have
expressed concern about the adequacy of protections under current
law for biometric data and an absence of clear criteria governing
data sharing.
-------------------------Indexing Terms-------------------------
REPORTNUM: GAO-03-174
ACCNO: A05491
TITLE: Technology Assessment: Using Biometrics for Border
Security
DATE: 11/14/2002
SUBJECT: Comparative analysis
Identity verification
Illegal aliens
Immigrants
National preparedness
Customs Service Advance Passenger
Information System
Customs Service/INS Interagency Border
Inspection System
Dept. of State Consular Lookout and
Support System
FBI Integrated Automated Fingerprint
Identification System
INS Passenger Accelerated Service System
INS Portable Automated Lookout System
Treasury Enforcement Communications
System
******************************************************************
** This file contains an ASCII representation of the text of a **
** GAO Product. **
** **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced. Tables are included, but **
** may not resemble those in the printed version. **
** **
** Please see the PDF (Portable Document Format) file, when **
** available, for a complete electronic file of the printed **
** document's contents. **
** **
******************************************************************
GAO-03-174
GAO
November 2002 GAO- 03- 174
United States General Accounting Office Technology Assessment Technology
Assessment
Using Biometrics for Border Security
Using Biometrics for Border Security
Page i GAO- 03- 174 Biometrics for Border Security Letter 1
Technology Assessment Overview 2 Purpose 2 Background 3 Results in Brief 4
Border Control Overview 7 Biometric Technologies 9 Scenarios for Using
Biometric Technologies for Border Security 11 The Role of Biometrics in
Border Security 16
Chapter 1 Introduction 18 The Federally Mandated Biometric Chimera System
19 An Overview of This Report 21
Chapter 2 Today*s U. S. Border Control Procedures 23 How U. S. Passports
Are Issued 24 How U. S. Visas Are Issued 29 Inspection at U. S. Ports of
Entry 32
Chapter 3 Biometric Technologies for Personal Identification 39 Biometrics
Defined 39 How the Technologies Work 39 Leading Biometric Technologies 45
Emerging Biometric Technologies 49 Common Applications of Biometric
Technologies 52 Performance Issues 54 Technologies Viable for U. S. Border
Control 69 Biometric Technology Applied to Border Control Today 74
Chapter 4 Scenarios for Border Control with Biometrics 79 Watch List Check
before Issuing Travel Documents 79 Watch List Check before Entering the
United States 83 U. S. Visas with Biometrics 85 U. S. Passports with
Biometrics 88 Implementing Multiple Scenarios 91 Contents
Page ii GAO- 03- 174 Biometrics for Border Security Chapter 5 Applying
Biometrics to Border Control: Challenges
and Implications 93 The Performance of Biometric Technologies 93 How
Introducing the Technology Affects People and Procedures 96 Weighing Costs
and Benefits 103 Effects on Privacy and the Economy 115
Chapter 6 Summary
121 Key Considerations in Using Biometrics for Border Control 122 High-
Level Analysis of Four Scenarios Using Biometrics 125 The Role of
Biometrics in Border Security 127 Agency Comments and Our Evaluation 129
External Reviewers* Comments 133
Appendix I Our Technology Assessment Methodology 136
Appendix II Fingerprint Recognition Technology 139 How the Technology
Works 142 The Leading Vendors 146 The Cost of Devices 147 Performance
Issues 147 User Acceptance 148 The Technology*s Maturity 149 Border
Control Applications Piloted and Deployed 157 Processing Issues 158 Device
Durability and Environmental Constraints 158
Appendix III Hand Geometry Technology 159 How the Technology Works 159 The
Leading Vendors 160 The Cost of Devices 160 Performance Issues 160 User
Acceptance 160 The Technology*s Maturity 161 Border Control Applications
Piloted and Deployed 163 Device Durability and Environmental Constraints
166
Page iii GAO- 03- 174 Biometrics for Border Security Appendix IV Facial
Recognition Technology 169
How the Technology Works 169 The Leading Vendors 172 The Cost of Devices
172 Performance Issues 173 User Acceptance 174 The Technology*s Maturity
175 Border Control Applications Piloted and Deployed 189 Processing Issues
191 Device Durability and Environmental Constraints 192
Appendix V Iris Recognition Technology 193 How the Technology Works 194
The Leading Vendors 196 The Cost of Devices 196 Performance Issues 196
User Acceptance 197 The Technology*s Maturity 197 Border Control
Applications Piloted and Deployed 199 Processing Issues 202 Device
Durability and Environmental Constraints 202
Appendix VI Cost Estimates for Using Biometrics for Border Security 203
Initial Cost Elements 203 Recurring Cost Elements 204 Assumptions 206
Estimated Costs for Conducting Watch List Checks with
Biometrics 207 Estimated Costs for Issuing Visas with Biometrics 208
Estimated Costs for Issuing Passports with Biometrics 215
Appendix VII Comments from the U. S. Department of State 222
Appendix VIII Comments from the U. S. Department of Justice 224
Page iv GAO- 03- 174 Biometrics for Border Security Appendix IX GAO
Contacts and Acknowledgments 229
GAO Contacts 229 Acknowledgments 229
Bibliography 230
Tables
Table 1: Leading Biometric Technologies 5 Table 2: Number of Inspections
at U. S. Ports of Entry, Fiscal Year
2001 9 Table 3: Estimated Costs for Implementing Border Security
Scenarios 15 Table 4: Number of Inspections at U. S. Ports of Entry,
Fiscal Year
2001 23 Table 5: Leading Biometric Technologies and Their Template Size 46
Table 6: Emerging Biometric Technologies and Their Maturity 50 Table 7:
Independent Biometric Test Results, 1991* 2002 60 Table 8: Four Viable
Biometric Technologies Compared 69 Table 9: The Enrollment Size of Seven
Operational Biometric
Systems 94 Table 10: Estimated Number of Biometric Matching Transactions
in Four Border Control Scenarios 94 Table 11: Security Risks and
Mitigating Techniques 103 Table 12: The Number and Type of Fraudulent
Documents INS
Inspectors Intercepted, Fiscal Year 2001 106 Table 13: Estimated Costs for
Watch List Checks 110 Table 14: Estimated Costs for Issuing Visas with
Biometrics 112 Table 15: Estimated Consular Costs for Issuing Visas with
Biometrics 112 Table 16: Estimated Costs for Issuing Passports with
Biometrics 113 Table 17: Cost Estimate Uncertainty Analysis for Four
Scenarios 114 Table 18: Summary of Biometric Systems Privacy Guidelines
118 Table 19: Estimated Costs for Implementing Border Security
Scenarios 125 Table 20: Leading Vendors of Fingerprint Recognition
Biometrics 147 Table 21: Summary of Results from the Fingerprint
Verification
Competition 2000 153 Table 22: Summary of Results from the Fingerprint
Verification
Competition 2002 154
Page v GAO- 03- 174 Biometrics for Border Security
Table 23: INS*s IDENT Fingerprint Benchmark Test Results, 1998 157 Table
24: Identix Airport Facial Biometric Pilot Results 176 Table 25: Facial
Recognition Product Usability Test 185 Table 26: Estimated Costs for Watch
List Checks before Issuing
Travel Documents and before Entering the United States 208 Table 27:
Estimated Costs for Issuing Visas with Biometrics Using
Fingerprint Recognition 210 Table 28: Estimated Costs for Issuing Visas
with Biometrics Using
Iris Recognition 211 Table 29: Estimated Costs for Issuing Visas with
Biometrics Using
Facial Recognition 212 Table 30: Estimated Costs for Issuing Visas with
Biometrics Using
Fingerprint and Iris Recognition 213 Table 31: Estimated Costs for Issuing
Visas with Biometrics Using
Fingerprint and Facial Recognition 214 Table 32: Estimated Costs for
Issuing Visas with Biometrics Using
Fingerprint, Iris, and Facial Recognition 215 Table 33: Estimated Costs
for Issuing Passports with Biometrics
Using Fingerprint Recognition 216 Table 34: Estimated Costs for Issuing
Passports with Biometrics
Using Iris Recognition 217 Table 35: Estimated Costs for Issuing Passports
with Biometrics
Using Facial Recognition 218 Table 36: Estimated Costs for Issuing
Passports with Biometrics
Using Fingerprint and Iris Recognition 219 Table 37: Estimated Costs for
Issuing Passports with Biometrics
Using Fingerprint and Facial Recognition 220 Table 38: Estimated Costs for
Issuing Passports with Biometrics
Using Fingerprint, Iris, and Facial Recognition 221
Figures
Figure 1: The U. S. Passport Application Process 25 Figure 2: A U. S.
Passport Cover 28 Figure 3: A U. S. Passport*s Biography Page 29 Figure 4:
The U. S. Visa Application Process 30 Figure 5: A U. S. Visa Foil 32
Figure 6: The U. S. Port of Entry Inspection Process 33 Figure 7: Motor
Vehicles Waiting for Inspection at the Paso del
Norte Port of Entry, El Paso, Texas 36 Figure 8: A Driver Being Questioned
at a Port of Entry 37 Figure 9: The Biometric Verification Process 42
Page vi GAO- 03- 174 Biometrics for Border Security
Figure 10: The Biometric Identification Process 44 Figure 11: The General
Relationship between FMR and FNMR 56 Figure 12: Standards for Biometric
Systems 66 Figure 13: The Front of a Laser Visa 76 Figure 14: The Back of
a Laser Visa 77 Figure 15: Issuing U. S. Visas by a Watch List Check
Process 80 Figure 16: Issuing U. S. Passports by a Watch List Check
Process 81 Figure 17: System Architecture for a Biometric Watch List Check
before Issuing Travel Documents 82 Figure 18: Entering the United States
by a Watch List Check
Process 83 Figure 19: System Architecture for a Biometric Watch List Check
before Entering the Country 84 Figure 20: Issuing U. S. Visas with
Biometrics 86 Figure 21: Port of Entry Visa Inspection with Biometrics 87
Figure 22: System Architecture for Issuing Visas with Biometrics 88 Figure
23: Issuing U. S. Passports with Biometrics 89 Figure 24: Port of Entry
Passport Inspection with Biometrics 90 Figure 25: System Architecture for
Issuing Passports with
Biometrics 91 Figure 26: Using Fingerprint Biometrics for Physical Access
140 Figure 27: Using Fingerprint Biometrics for Logical Access 140 Figure
28: A Fingerprint Biometric Device for Personal
Identification 141 Figure 29: Common Fingerprint Features 144 Figure 30:
Established Fingerprint Types. 145 Figure 31: An IDENT Workstation 150
Figure 32: Fingers Guided by Pegs in a Biometric Hand Geometry
Measurement 159 Figure 33: A Traveler Using an INSPASS Hand Geometry
Device 164 Figure 34: A Traveler Using Ben Gurion Airport*s Biometric Hand
Geometry System 165 Figure 35: A Typical Hand Geometry Recognition Device
167 Figure 36: A Hand Geometry Recognition Device That Is Enclosed 168
Figure 37: Local Feature Analysis: A Topographical Grid of Facial
Regions 170 Figure 38: Two- Dimensional, Gray- Scale Images of an
Eigenface
Template 171 Figure 39: CCTV Surveillance Equipment 173 Figure 40: Facial
Recognition Distance Identification 178 Figure 41: Facial Recognition
Distance Verification 179 Figure 42: Facial Recognition Expression
Identification 180
Page vii GAO- 03- 174 Biometrics for Border Security
Figure 43: Facial Recognition Expression Verification 180 Figure 44:
Facial Recognition Media Identification: Digital to
35 mm 181 Figure 45: Facial Recognition Media Verification: Digital to 35
mm 182 Figure 46: Facial Recognition Pose Identification 183 Figure 47:
Facial Recognition Temporal Identification 184 Figure 48: Facial
Recognition Temporal Verification 184 Figure 49: The Iris and Other Parts
of the Eye 193 Figure 50: Iris Recognition Physical Access Control System
194 Figure 51: Iris Recognition System with Desktop Camera 195 Figure 52:
Mapping the Eye for Iris Recognition Systems 196 Figure 53: Iris
Recognition Device for Border Control at London*s
Heathrow Airport 200 Figure 54: Border Control Lane with Iris Recognition
Device at
London*s Heathrow Airport 200
Abbreviations
AAMVA American Association for Motor Vehicle Administration AFIS automated
fingerprint identification system ANSI American National Standards
Institute API application programming interface APIS Advance Passenger
Information System ATM automated teller machine BAPI biometric application
programming interface CBEFF Common Biometric Exchange File Format CCD
Consular Consolidated Database CCTV closed- circuit television CLASS
Consular Lookout and Support System DOD Department of Defense EER equal
error rate EPPS Expedited Passenger Processing System ETT enrollment timed
test FAA Federal Aviation Administration FBI Federal Bureau of
Investigation FERET Face Recognition Technology FMR false match rate FNMR
false nonmatch rate FRVT 2000 Facial Recognition Vendor Test 2000 FRVT
2002 Facial Recognition Vendor Test 2002 FTE failure to enroll
Page viii GAO- 03- 174 Biometrics for Border Security
FTER failure to enroll rate FVC 2000 Fingerprint Verification Competition
2000 FVC 2002 Fingerprint Verification Competition 2002 IAFIS Integrated
Automated Fingerprint Identification System IAFS Immigration and Asylum
Fingerprint System IBG International Biometric Group IBIA International
Biometric Industry Association IBIS Interagency Border Inspection System
ICAO International Civil Aviation Organization IDENT Automated Biometric
Fingerprint Identification System INCITS InterNational Committee for
Information Technology
Standards INS Immigration and Naturalization Service INSPASS INS Passenger
Accelerated Service System IRS Internal Revenue Service JPEG Joint
Photographic Experts Group LFA local feature analysis NAFTA North American
Free Trade Agreement NAS National Academy of Sciences NIST National
Institute of Standards and Technology NPL National Physical Laboratory NSA
National Security Agency OIDTT old image database timed test PALS Portable
Automated Lookout System PFM Passport Files Miniaturization PIN personal
identification number PRISM Passport Records Imaging System Management RSI
Recognition Systems Inc. SENTRI Secure Electronic Network for Travelers
Rapid Inspection TECS Treasury Enforcement Communications System WSQ
wavelet scalar quantization
Page 1 GAO- 03- 174 Biometrics for Border Security
November 15, 2002 The Honorable Richard J. Durbin Chairman The Honorable
Robert F. Bennett Ranking Minority Member Subcommittee on Legislative
Branch Committee on Appropriations United States Senate
As directed in the Fiscal Year 2002 Legislative Branch Appropriations
Conference Report (House Report 107- 259) and subsequent support letters
from interested Members of the Congress, we conducted a pilot program in
technology assessment that examined the use of biometric technologies for
border control. This report discusses the current maturity of several
biometric technologies and possible implementation of these technologies
in current border control processes. Policy implications and key
considerations for the use of biometric technologies are also discussed.
We are sending copies of this report to the Attorney General, the
Secretary of State, and interested congressional committees. We will
provide copies to others on request. In addition, the report will be
available at no charge on the GAO Web site at http:// www. gao. gov.
If you have questions concerning this report, you may contact me on (202)
512- 2700 (kingsburyn@ gao. gov) or Naba Barkakati, Senior Level
Technologist, on (202) 512- 4499 (barkakatin@ gao. gov). Major
contributors to this report are listed in appendix IX.
Nancy R. Kingsbury Managing Director Applied Research and Methods
United States General Accounting Office Washington, DC 20548
Technology Assessment Overview Page 2 GAO- 03- 174 Biometrics for Border
Security
One facet of the homeland security strategy focuses on border security*
preventing the illegal entry of people and goods into the United States
without impeding their legitimate flow. Security concerns need to be
balanced with practical cost and operational considerations as well as
political and economic interests. A risk- based approach can help identify
and address security concerns. This is a challenging mission because
the nation shares a 5,525 mile border with Canada and a 1,989 mile
border with Mexico and has a shoreline of about 95, 000 miles,
there are almost 400 official entry points along these borders, and
there were more than 500 million border crossings into the United States
last year, two- thirds by travelers who were not citizens.
Part of the border security mission is controlling the passage of
travelers through these official entry points into the United States.
Biometric technologies, using one or more of a person*s distinct
physiological or behavioral characteristics, have been suggested as a way
to help automate the identification of travelers to the United States at
these ports of entry.
As directed in the Fiscal Year 2002 Legislative Branch Appropriations
Conference Report (House Report 107- 259) and subsequent support letters
from interested Members of the Congress, this technology assessment
focuses on four key questions:
1. What biometric technologies are currently deployed, currently available
but not yet deployed, or in development that could be deployed in the
foreseeable future for use in securing the nation*s borders?
2. How effective are these technologies now or likely to be in the future
in helping provide security to the nation*s borders?
3. What are the economic and effectiveness trade- offs of implementing
these technologies?
4. What are the implications of using biometric technologies for personal
security and the preservation of individual liberties?
To answer these questions, we convened, with the assistance of the
National Academy of Sciences, two meetings on biometrics and border
control issues that included manufacturers of facial, fingerprint, and
iris recognition and hand geometry technologies, as well as informed
Technology Assessment Overview
Purpose
Technology Assessment Overview Page 3 GAO- 03- 174 Biometrics for Border
Security
representatives from academia, government, and industry groups; privacy
and civil liberty advocates; and other stakeholders such as
representatives of border communities and trade organizations. We also
interviewed certain users of biometric technologies, including the Federal
Bureau of Investigation, Immigration and Naturalization Service (INS),
National Security Agency, National Institute of Standards and Technology,
the Department of State, and the Canada Customs and Revenue Agency. We
reviewed test documentation to understand the performance of biometric
technologies and visited a number of ports of entry where these
technologies may be used. We interviewed manufacturers of biometric
technologies and reviewed their publications to obtain descriptive
information about their equipment. We interviewed officials from biometric
industry organizations, including the Biometric Consortium and the
Biometric Foundation. We also interviewed the International Biometric
Group (IBG). We postulated four scenarios for using biometric technologies
in border security and created cost models to estimate the rough order of
magnitude costs of implementing biometric technologies. We provided our
assessment report to the Department of Justice and the Department of State
for their review. We also had the draft report reviewed by a number of
external experts.
Our report starts with a description of the current border control
procedures for admitting people into the United States* issuing visas to
citizens of other nations and passports to U. S. citizens and inspecting
travelers at the ports of entry. Next, the report describes how biometric
technologies work, including the different types of biometric
technologies, their levels of maturity, and their operating and
performance characteristics. We present four possible scenarios in which
biometrics might be applied to current U. S. border control procedures.
For each scenario, we analyze some of the costs, benefits, and risks
associated with implementation. Finally, the report sums up certain policy
implications and challenges to be faced if a biometric system is to be
designed and deployed for border security. A number of appendixes provide
details on the major biometric technologies.
The United States essentially relies on a two- step approach to prevent
inadmissible people from entering the country. The Bureau of Consular
Affairs in the State Department is responsible for issuing international
travel documents, such as passports in the United States and visas in
other countries, and INS in the Department of Justice is responsible for
inspecting travelers at the ports of entry. Background
Technology Assessment Overview Page 4 GAO- 03- 174 Biometrics for Border
Security
The term biometrics covers a wide range of technologies that can be used
to verify a person*s identity by measuring and analyzing his or her
characteristics. Identifying a person*s physiological characteristics is
based on data derived from measuring a part of the body directly.
Technologies have been developed to measure people*s fingers, hands,
faces, and eye retinas and irises. Identifying a person*s behavioral
characteristics is based on data derived from an individual*s actions,
such as how he or she talks, types, or signs his or her name. Biometric
systems are essentially pattern recognition systems. They use electronic
or optical sensors such as cameras and scanning devices to capture images,
recordings, or measurements of a person*s characteristics and computer
hardware and software to extract, encode, store, and compare these
characteristics.
Using biometrics as identifiers for border security purposes appears to be
appealing because they can help tightly bind a traveler to his or her
identity by using physiological or behavioral characteristics. Unlike
other identification methods, such as identification cards or passwords,
biometrics are less easily lost, stolen, or guessed.
Biometrics have been implemented to a limited degree in U. S. border
control systems. For example, since 1993, the INS Passenger Accelerated
Service System (INSPASS) has allowed for automated inspections of more
than 35,000 frequent fliers at nine airports. The Congress has enacted
laws in the past 6 years that require a more extensive use of biometrics
in border control systems. These laws require that by the end of 2004, all
ports of entry are to be able to perform biometric comparison and
authentication of all U. S. visas and other travel and entry documents and
that all systems of the State Department, INS, and federal law enforcement
and intelligence agencies that contain information about aliens are to be
interoperable.
Biometric technologies are available today and are being used for a
variety of applications such as access control and criminal identification
and surveillance. We considered a number of leading and emerging biometric
technologies that could potentially be used for securing the nation*s
borders. The seven leading biometric technologies include facial
recognition, fingerprint recognition, hand geometry, iris recognition,
retina recognition, signature recognition, and speaker recognition (see
table 1). Of these, fingerprint recognition, facial recognition, iris
recognition, and hand geometry appeared to be suitable for border security
because all have been used in border control pilots and applications.
However, hand geometry is not highly distinctive and cannot reliably pick
out an Results in Brief
Technology Assessment Overview Page 5 GAO- 03- 174 Biometrics for Border
Security
individual from among many. Consequently, hand geometry is not suitable if
there is a need to search the biometrics database to determine if a person
has previously enrolled in the database or is in a watch list. However,
hand geometry is viable for verifying claimed identity when another
biometric technology is used for the identification check during
enrollment. We also looked at emerging biometric technologies, such as ear
shape recognition and odor sensing, and found that they are in various
stages of development and have not yet been used in border control
applications. Our assessment is based on a snapshot of biometric
technologies as they existed in early 2002.
Table 1: Leading Biometric Technologies Technology How it works Suitable
for
border control
Facial recognition Captures and compares facial patterns Yes Fingerprint
recognition Captures and compares fingertip patterns Yes Hand geometry
Measures and compares dimensions of hand and fingers Yes (verification
only) Iris recognition Captures and compares iris patterns Yes Retina
recognition Captures and compares retina patterns No Signature recognition
Captures and compares rhythm, acceleration, and pressure flow of signature
No Speaker recognition Captures and compares cadence, pitch, and tone of
vocal tract No
Source: GAO analysis.
To evaluate the effectiveness of biometrics in border control, it is
important to recognize that the use of biometric technology would be but
one component of the decision to support systems that determine who is
allowed to enter the United States and who is not. Biometric technology
can play a role in associating a person with travel documents such as
visas and passports. When used at a border inspection, the biometric
comparison can be used to help decide whether to admit a traveler into the
United States.
When biometric technology is used in border control, the border control
processes will have to be changed not only to use the new technology but
also to compensate for its shortcomings. None of these technologies have
been used in an application as large as that required for a border control
system. Further, biometric technologies are not perfect* all have some
measured rates of erroneously matching a person or erroneously not
matching a person. The people involved, such as travelers, inspectors, and
consular personnel, will have to be trained in how to use the new system
and in the new border control processes.
Technology Assessment Overview Page 6 GAO- 03- 174 Biometrics for Border
Security
Before any decision is made to implement biometrics in a border control
system, the benefits of the system must be weighed against its costs. The
purpose of any biometrics initiative is to prevent the entry of travelers
who are inadmissible to the United States. For example, using a biometric
watch list can provide an additional check to name- based checks and can
help detect travelers trying to evade detection who have successfully
established a separate name and identity. The use of passports and visas
with biometrics can help positively identify travelers as they enter the
United States and can limit the use of fraudulent documents, including
counterfeit and modified documents, and impostors* use of legitimate
documents.
To analyze the costs of using three biometric technologies* facial,
fingerprint, and iris recognition* we define four scenarios in which these
technologies can be used to support border control operations. Two
scenarios use a biometric watch list to identify travelers who are
inadmissible to the United States (1) before issuing travel documents or
(2) before travelers enter the country. To help bind the claimed identity
of travelers to their travel documents, biometrics could be incorporated
into (1) U. S. visas or (2) U. S. passports. As defined, these four
scenarios are not mutually exclusive and could be implemented
independently or in combination. The costs of a biometric border control
system will not be trivial. For example, our rough order of magnitude cost
estimates to implement visas with biometrics are between $1.3 billion and
$2.9 billion initially and between $0.7 and $1.5 billion annually
thereafter.
Finally, important policy implications must be addressed in trade- offs
between increasing security and the impact on areas such as privacy,
economy, traveler convenience, and international relations. Civil
liberties groups and privacy experts have expressed concern about the
adequacy of protections under current law for biometric data and an
absence of clear criteria governing data sharing. Requiring biometric-
enabled visas could potentially affect the travel and tourism industry
adversely. Increased inspection times because of biometric identification
checks could result in longer waiting times, especially at land crossings,
causing local merchants on both sides of the border to lose sales.
International relations could be affected as other countries reciprocate
when the United States asks visitors from those countries to provide
biometric identifiers when they apply for visas.
Whether the financial and nonfinancial costs are warranted by the benefits
of greater security is a policy issue that should be determined before
biometric technologies are implemented in a border control system. This
Technology Assessment Overview Page 7 GAO- 03- 174 Biometrics for Border
Security
report provides useful information that can help serve as the basis for
these decisions. As our report describes, biometric technology is not a
panacea for the border security problem. It is only one component of the
decision support systems that determine who is allowed to enter the United
States and who is not. A risk- based approach would be helpful in
addressing the overall border security problem and the high- level goals
that can be achieved with biometric technologies. The approach could rely
on establishing what is being protected, who the adversaries are, what the
vulnerabilities are, what the priorities are, and what mitigation
strategies can be implemented. Answering these questions should help
determine the proper role of biometric technologies in border security.
We provided a draft of this report to the Department of Justice and the
Department of State for their review. The Department of Justice expressed
some concerns, but the State Department stated that it appreciated the
thorough and balanced approach we took in our assessment of the use of
biometrics for border security. We include State*s and Justice*s comments
in appendixes VII and VIII, respectively, and summarize them in chapter 6.
State and Justice also provided technical comments on the draft, which we
incorporated as appropriate.
We also provided a draft of this report to 16 different organizations,
representing government, industry, and academia, for their review. We
received comments and suggestions from 10 reviewers. The comments included
the correction of technical inaccuracies and the highlighting of certain
aspects of the assessment that reviewers considered important. We have
incorporated these comments, where appropriate, in the report. We
summarize these comments in chapter 6.
The United States relies essentially on two primary procedures to
facilitate the entry of people authorized to enter the country and to
ensure that inadmissible people are prevented from entering. The State
Department*s Bureau of Consular Affairs issues international travel
documents, including passports to U. S citizens and visas to people who
are not U. S. citizens and are traveling to the United States. INS
inspects travelers entering the United States through official ports of
entry. In addition, INS*s Border Patrol is responsible for securing the
borders and apprehending travelers entering through other than official
ports of entry.
Passports are issued to U. S. citizens to permit their travel abroad and
to facilitate their entry back into the United States. U. S. citizens can
apply for Border Control
Overview Passport Processing
Technology Assessment Overview Page 8 GAO- 03- 174 Biometrics for Border
Security
passports at one of more than 4,500 passport acceptance offices. Few of
these offices are State Department offices* most are offices in facilities
such as U. S. post offices or state, county, township, and municipal
government offices. Passport acceptance agents review application packages
for completeness and complete a checklist regarding their impressions of
applicants and their applications. After the applications are sent to the
central application processing center, they are run through a State
Department computer system that checks to see (1) whether the applicant
has been identified as someone who is not eligible to receive a passport,
(2) whether the individual already has an active passport, and (3) whether
the individual has multiple applications in process. Passport examiners
review the results of these checks and the applications and decide whether
to issue passports. If an application is approved, a passport is generated
and sent to the applicant.
With some exceptions, visitors to the United States are required to have a
visa to enter. Worldwide, travelers can apply for a visa at 210 embassies
and consulates. Visa applications are entered into a State Department
computer system and are checked to determine items such as whether an
applicant has been identified as someone who is not eligible to receive a
visa, whether the applicant*s passport matches a passport that has been
reported as lost or stolen, or whether the applicant has been refused a
visa in the past. In some cases, an interview with the visa applicant or a
security advisory opinion from State headquarters is required. In
determining whether to grant the visa, the consular officer reviews the
data provided in the application and the computer system and, if
applicable, the interview and security advisory opinion. If the
application is approved, a visa foil is generated and provided to the
traveler.
All people legally entering the United States must be processed through an
air, land, or sea port of entry. As shown in table 2, about 82 percent of
border crossings occurred at land ports of entry last year. An individual
entering the country through an official port of entry first enters a
process called primary inspection. Inspectors determine whether travelers
qualify for admission or additional review is necessary. If additional
review is necessary, the individual is referred to secondary inspection,
where a final decision on whether to admit the traveler is made. During
fiscal year 2001, about 1.7 percent of the more than 500 million border
crossers entering the country were referred to secondary inspection, where
707,920 were denied admission. Visa Processing
Port of Entry Inspections
Technology Assessment Overview Page 9 GAO- 03- 174 Biometrics for Border
Security
Table 2: Number of Inspections at U. S. Ports of Entry, Fiscal Year 2001
Type of port Number of ports Number of inspections
Sea 86 11,952,501 Air 155 79,598,681 Land 154 414,364,965
Total 395 505,916,147
Source: GAO analysis of INS data.
The processes used for primary inspection vary, depending on the mode of
travel* air, land, or sea* and the traveler*s nationality. INS uses a
combination of methods to inspect travelers, including a brief interview
with the travelers, an inspection of their travel or identification
documents, and computer checks of their names or the license plates of
their vehicles. The traveler*s nationality also dictates the documentation
requirements. For example, U. S. citizens do not require passports unless
they are returning from outside the Western Hemisphere. In general, aliens
must present their passport and a U. S.- issued visa. Citizens of
countries participating in the visa waiver program do not require a visa
to enter the United States.
Biometric technologies have been used in a wide array of applications,
including access control to buildings and computers, criminal
identification and surveillance, licensing and voter applications, and
fraud reduction. Biometric technologies can be used in a verification or
identification mode. Regardless of the method used, an enrollment process
is required to capture a biometric sample, extract and encode the sample
as a biometric template, and store the data in a database for future
comparisons. In verification mode (e. g., access control to a building
with an identity card), the biometric system verifies the validity of a
claimed identity, answering the question *Is this person who she claims to
be?* In identification mode (e. g., criminal surveillance), the biometric
system compares the individual*s biometric with all stored biometric
records to answer the question *Who is this person?*
We considered seven leading biometric technologies: facial recognition,
fingerprint recognition, hand geometry, iris recognition, retina
recognition, signature recognition, and speaker recognition. Four* facial
recognition, fingerprint recognition, hand geometry, and iris recognition*
appear to be suitable for border control applications. All four have been
used in border control pilots and applications. The three other
technologies have key Biometric
Technologies
Technology Assessment Overview Page 10 GAO- 03- 174 Biometrics for Border
Security
problems that inhibit their use for border control. Retina recognition is
considered to be too intrusive because many users experience discomfort in
using the devices, which operate close to their eyes. Signature
recognition has a high error rate because it has been found that people do
not always sign their name the same way each time. Speaker recognition has
been piloted in a border control environment but has been found to be
unreliable. Also, speaker recognition systems do not perform well in noisy
environments such as would be encountered at ports of entry.
The emerging technologies we considered* facial thermography, gait
recognition, ear shape recognition, DNA matching, odor sensing, blood
pulse measurement, skin pattern recognition, vein scan, and nailbed
identification* are in various stages of development and have not yet been
used in border control applications.
Fingerprint recognition has been widely used and accepted, primarily in
law enforcement, for four decades. Facial recognition can be used to
compare either a live facial scan to a stored biometric template or a
static image to a digitized photograph. Facial images are already
prevalent in travel documents, and people are accustomed to having their
picture taken. Hand geometry has been widely used in access control
applications and is relatively easy to use. Iris recognition identifies
people by numerous characteristics of the colored ring surrounding the
pupil of the eye, some of which tend to remain stable throughout life.
In order to differentiate between biometric technology products, they are
often characterized by factors such as accuracy, testing, standards, and
user acceptance. The accuracy of a biometric technology is usually
measured by three key error statistics: the rate at which a system
erroneously matches a person, the rate at which a system erroneously does
not match a person, and the rate at which people are unable to enroll in a
system. To evaluate biometric technologies, the results of independent
tests should be consulted. In addition, tests have been conducted in which
researchers have successfully fooled biometric systems with artificial
characteristics such as a latex finger or a facial picture. Adherence to
standards enhances the ability of a biometric device to store and exchange
data. Another factor to consider in selecting a biometric technology is
the ease of use. Some people find biometric technologies difficult, if not
impossible, to use. Still others resist biometrics in general as
intrusive, inherently offensive, or just uncomfortable to use.
No biometric technology is best for every situation, but it is possible to
determine the most accurate, easiest to use or deploy, or cheapest,
Technology Assessment Overview Page 11 GAO- 03- 174 Biometrics for Border
Security
depending on the objectives to be achieved. For example, hand geometry
requires the least data storage, fingerprint and iris recognition have the
lowest error rates, and facial recognition is the easiest to use. However,
each technology also has its limitations. For example, about 2 to 5
percent of people cannot be easily fingerprinted because their
fingerprints have become dry or worn from age, extensive manual labor, or
exposure to corrosive chemicals. Facial recognition systems have not
performed particularly well in independent testing. Iris recognition is a
relatively new technology and has not been used in any large operational
applications as fingerprint and facial recognition systems have. Hand
geometry is not highly distinctive and thus is not suitable for
identification applications. These limitations and others would have to be
considered if these technologies were to be deployed within a border
control system. (More details on the biometric technologies can be found
in chapter 3 and appendixes II to V.)
We developed and analyzed four different scenarios in which fingerprint,
facial, or iris recognition biometric technologies or some combination of
them could be used to improve current border control procedures. Two
scenarios use a biometric watch list to identify travelers who are
inadmissible to the United States (1) before issuing travel documents or
(2) before travelers enter the country. To help bind travelers to their
travel document, two other scenarios could be used to incorporate
biometrics into (1) U. S. visas or (2) U. S. passports. These four
scenarios can be implemented independently or in combination.
The first scenario involves the use of facial recognition to help identify
people ineligible to receive a passport or a visa. The biometric
identification check would be conducted at the same time as other computer
checks are conducted on each travel document application. The second
scenario uses an automated facial recognition system at the ports of entry
that can observe a person*s face and check the observed facial features
against a watch list of people who should be denied access to the country.
Both scenarios require the creation of a biometric- based watch list that
stores photographs of individuals selected according to criteria
determined by border security and other law enforcement agencies. While
both scenarios require a centralized facial recognition server to perform
matches, performing checks at the ports of entry would also require the
purchase of facial recognition systems for the almost 4,000 inspection
stations at the ports of entry.
The two other scenarios introduce biometrics to visas and passports. In
both of these scenarios, travel document applicants would be required to
Scenarios for Using
Biometric Technologies for Border Security
Technology Assessment Overview Page 12 GAO- 03- 174 Biometrics for Border
Security
have their biometric sample collected* at 1 of 210 embassies and
consulates for visa applicants or at 1 of 4,500 passport acceptance
offices for passport applicants. As part of the enrollment and document
issuance process, an additional identification check of applicants would
be made against the database of issued documents to ensure that a person
does not receive multiple documents under different identities. Biometric
scanners would also have to be installed at the ports of entry to verify
the identity of travelers with biometrically enabled travel documents.
The successful implementation of any technology depends not only on the
performance of the technology but also on the operational processes that
employ the technology and the people who execute them. The implementation
of biometrics in border security is no exception. Further, the use of
technology alone is not a panacea for the border security problem.
Instead, biometric technology is just a piece of the overall decision
support system that helps determine whether to allow a person into the
United States. The first decision is whether to issue travelers a U. S.
travel document. The second decision, made at the ports of entry, is
whether to admit travelers into the country. Biometrics can play a role in
both decisions. Sorting the admissible travelers from inadmissible ones is
now done by using information systems for checking names against watch
lists and by using manual human recognition capabilities to see if the
photograph on a travel document matches the person who seeks entry to the
United States. When enabled with biometrics, automated systems can verify
the identity of the traveler and assist inspectors in their decision
making.
The four biometric scenarios will affect key border security processes. A
key factor is the performance of the biometric technology. For example, if
the biometric technology that is used to perform watch list checks before
travel documents are issued has a high rate of false matches, workload
could increase at the embassies and consulates for visas and at the
passport centers for passports. If the same biometric solution were used
at the ports of entry, it could lead to increased delays in the inspection
process and an increase in the number of secondary inspections.
Exception processing will have to be carefully considered. Exceptions
include people who fail to enroll in a system or are not correctly matched
by a verification system. Exception processing that is not as good as
biometric- based primary processing could be exploited as a security hole.
Failure of equipment must also be considered and planned for. Further, for
issuing visas or passports with biometrics, an appropriate transition The
Effect on Border
Control Processes
Technology Assessment Overview Page 13 GAO- 03- 174 Biometrics for Border
Security
strategy must be devised to simultaneously handle biometric travel
documents and the current travel documents that could remain valid without
biometrics for the next 10 years.
Implementing biometrically enabled travel documents requires a strong
binding and verification process to tie individuals to their identities
using their biometrics. A process that does not have strong binding
mechanisms can provide little improvement over existing procedures. A
failure in the enrollment or the verification process could undermine the
use of biometric technologies. For example, procedures must be developed
to handle individuals who could not be enrolled in the system. Even if
individuals are properly enrolled, they might not be properly matched
during inspection. Adequate procedures have to be in place to properly
differentiate between system problems and persons who are impostors or
otherwise inadmissible to the United States. Information security also is
important in ensuring strong binding. If rogue individuals can modify the
biometric database or the token on which individual biometric records are
stored, a person*s bond to his or her biometric data can be compromised.
Before any significant project investment is made, the benefit and cost
information of the project should be analyzed and assessed in detail. The
project concept should be based on high- level system goals, which for a
border control system would include items such as binding a biometric
feature to a person*s identity on a travel document, identifying
undesirable persons on a watch list, checking for duplicate enrollments in
the system, verifying identities at the borders, ensuring the security of
the biometric data, and ensuring the adequacy of privacy protections.
The desired benefit of all the scenarios we describe* the use of biometric
watch lists or biometrically enabled travel documents* is the prevention
of the entry of travelers who are inadmissible to the United States. More
specifically, the use of a biometric watch list can provide an additional
check to name- based checks and can help detect travelers who are trying
to evade detection and have successfully established separate names and
identities. The use of passports and visas with biometrics can help
positively identify travelers as they enter the United States and can
limit the use of fraudulent documents, including counterfeit and modified
documents and impostors* use of legitimate documents.
These benefits have several limitations. First, the benefit achieved in
each scenario is directly related to the performance of the biometric
Maintaining Information
Security Weighing Costs and Benefits
Technology Assessment Overview Page 14 GAO- 03- 174 Biometrics for Border
Security
technology. The performance of facial, fingerprint, and iris recognition
is unknown for systems as large as a biometric visa system that would
require the storage and comparison against 100 million to 240 million
records. The largest facial, fingerprint, and iris recognition systems
contain 60 million, 40 million, and 30,000 records, respectively.
For the watch list scenarios, the population of the watch list is critical
to the system*s effectiveness. Issuing passports and visas with biometrics
will only assist in identifying those currently required to obtain
passports or visas to enter this country. For example, U. S. citizens do
not have to have a passport to return from Canada or Mexico. Canadians,
Mexicans with border crossing cards, and aliens participating in the visa
waiver program do not have to have a visa to enter the United States. The
issuance of passports and visas with biometrics is also dependent on
establishing the correct identity during enrollment. This process will
typically be dependent on the presentation of identification documents. If
the documents do not specify the applicant*s true identity, then the
travel document will still be linked to a false identity.
Further, biometric technology is not a solution to all border security
problems. Biometric technology can address only problems associated with
identifying travelers at official locations such as embassies, passport
acceptance offices, and ports of entry. While the technology can help
reduce the number of illegal immigrants who cross with fraudulent
documents, it cannot help with illegal immigrants who cross *between the
borders* and not at a port of entry. INS has previously estimated that up
to 60 percent of the 275,000 new illegal immigrants a year do not present
themselves at a port of entry to enter the United States. In addition,
biometrics cannot help with aliens who enter through ports of entry and
are properly admitted by an inspector but may overstay their visit.
The security benefits gained from the use of biometrics must be weighed
against the cost of implementing the scenario. For each of the four
scenarios, we created cost models to estimate the cost of developing,
implementing, and maintaining various biometric processes. We included the
costs of both the technology and the effects on people and processes.
Table 3 summarizes the initial and annual recurring costs of implementing
each scenario. The initial costs include elements such as development,
installation, training, biometric hardware and software, and consular
facility renovation. The recurring costs include elements such as
biometric hardware and software maintenance, system support and
operational personnel, consular personnel, facility maintenance, and
annual supplies. While the costs of people and space required to enroll
travelers in
Technology Assessment Overview Page 15 GAO- 03- 174 Biometrics for Border
Security
biometric systems at embassies and consulates are included, the costs of
people and space required to verify the biometrics at ports of entry are
not included. Consular staff and space are major cost drivers. For
example, for issuing visas with biometrics, these costs make up between 21
percent and 31 percent of the system*s total initial cost and between 23
percent and 29 percent of its total recurring cost.
Table 3: Estimated Costs for Implementing Border Security Scenarios
Scenario Initial
cost Annual
recurring cost
Watch list check before issuing travel documents $53 $73 Watch list check
before entering the United States 330 237 Issuing visas with biometrics
1,399* 2,845 698* 1, 482 Issuing passports with biometrics 4,446* 8,766 1,
555* 2,363
Note: Dollars are in millions. Source: GAO analysis.
The watch list scenarios assume the use of facial recognition technology,
because faces from photographs are often the only biometric available for
individuals who may be inadmissible to the United States. Travel documents
with biometrics can use facial, fingerprint, or iris recognition or some
combination of the three.
The Privacy Act of 1974 limits federal agencies* collection, use, and
disclosure of personal information, including personal information such as
finger or voice print and photographs. Accordingly, the Privacy Act
generally covers federal agency use of personal biometric information.
However, as a practical matter, the act is likely to have a more limited
application for border security. First, the act applies only to U. S.
citizens and lawfully admitted permanent resident aliens. Second, the act
includes exemptions for law enforcement and national security purposes.
Representatives of civil liberties groups and privacy experts have
expressed concerns regarding (1) the adequacy of protections for security,
data sharing, identity theft, and other identified uses and (2) secondary
uses and *function creep.* The Internal Revenue Service, the RAND
Corporation, and IBG have developed privacy frameworks that establish
guidelines on issues with the scope and capabilities of biometric systems,
the protection of data, the protection of users, and the disclosure,
auditing, accountability, and oversight of biometric systems. Protecting
Privacy and
Civil Liberties
Technology Assessment Overview Page 16 GAO- 03- 174 Biometrics for Border
Security
Any lengthening in the process of obtaining travel documents or entering
the United States could affect travelers significantly. At some consular
posts, visas are issued the day applications are received. Even without
biometrics, the busiest ports of entry regularly have delays of 2 to 3
hours. Increases in inspection times could compound these delays. Delays
inconvenience travelers and could result in fewer visits to the United
States or lost business to the nation. Further studies will be necessary
to measure what the potential effect could be on the American economy and,
in particular, on the border communities. These communities depend on
trade with Canada and Mexico, which totaled $653 billion in 2000.
Finally, the use of biometrics in the United States could affect the
number of international visitors and how other countries treat visitors
from the United States. Visitors from some countries may not want to come
to the United States if it is less convenient to do so. In addition,
because much of visa issuance policy is based on reciprocity* the process
for allowing a nation*s citizens to enter the United States is similar to
the process followed by that nation for visitors from the United States*
other nations may start requiring biometric samples from U. S. citizens if
the United States requires biometric samples from their citizens. (More
details on costs and benefits, as well as the potential implications, of
using biometrics are provided in chapter 5.)
People are identified by three basic means: by something they know,
something they have, or something they are. Current U. S. border security
processes identify travelers by using travel documents such as passports
and visas and asking travelers questions* things the travelers have and
know. The travel document also establishes a traveler*s eligibility to
enter the country.
The use of biometrics* things the travelers are* can more securely bind a
person*s identity to a travel document. Two processes are keys to
achieving this binding. First, a strict and thorough enrollment step is
necessary to bind a person to an identity. The identity claimed by the
traveler is based on documents such as a birth certificate, passport, or
other government- issued documents. If processes are not in place to
ensure the validity of the traveler*s claimed identity, the person could
be linked with a false identity. Second, an effective matching process is
required to link the person to the travel document. If a person can bypass
the biometric check or can deceive the biometric system, the person may be
erroneously granted admission to the United States. The performance of the
biometric technology is also important to the execution of these The
Effect on
Convenience, the Economy, and International Relations
The Role of Biometrics in Border Security
Technology Assessment Overview Page 17 GAO- 03- 174 Biometrics for Border
Security
processes. Effective enrollment and matching processes could allow for the
use of biometric- enabled travel documents that will establish not only
the traveler*s eligibility to enter the country but also that the traveler
is indeed the individual depicted on the document.
However, biometric technology is just one component of the decision
support systems that help determine who is allowed to enter the United
States and who is not. For example, the technologies may be able to reduce
document fraud but may not be able to detect illegal entry to the United
States through other than official ports of entry. A risk- based approach
would be helpful in addressing the overall border security problem and the
high- level goals that can be achieved with biometric technologies. The
approach could rely on answering five basic questions: What are we
protecting? Who are the adversaries? What are the vulnerabilities? What
are the priorities? What mitigation strategies can be used? A decision to
implement our four scenarios or any others should be based on an approach
that answers these questions. The scenarios could be partially implemented
or combined in different ways. New scenarios could be defined in which
travelers voluntarily enroll in a biometric identification system similar
to INSPASS for expedited border crossing. Trade- offs should be made to
determine the best implementation of biometrics for border security. For
example, a partial implementation may be less costly without sacrificing
any of the security benefits.
Regardless of how biometric technology is used in border security, using a
risk- based approach should help in developing the high- level goals of a
system and its concept of operation. The answers should also help point
out the limitations of such a system and what it will not be able to
provide. They can also play a role in the analysis and weighting of the
benefits in a cost- benefit analysis, as well as the trade- off analysis
between greater security and issues such as privacy and the economy. With
these answers, the proper role of biometric technology in border security
can be determined.
Chapter 1: Introduction Page 18 GAO- 03- 174 Biometrics for Border
Security
A primary element of the homeland security strategy is the improvement of
U. S. border security* preventing the illegal entry of people and goods
into the United States while facilitating their legitimate flow. Security
concerns need to be balanced with practical cost and operational
considerations as well as political and economic interests. The United
States shares a 5, 525 mile border with Canada and a 1,989 mile border
with Mexico. Its maritime border includes 95,000 miles of shoreline. There
were more than half a billion border crossings into the United States last
year; about two- thirds were not by U. S. citizens. The number of distinct
travelers into the country each year is unknown because some people enter
the country many times in one year, some daily.
Facilitating the flow of people while preventing illegal border crossings
is a matter of identifying travelers. People are identified by three basic
means: by something they know, something they have, or something they are.
People and systems regularly use these means to identify people in
everyday life. For example, members of a community routinely recognize one
another by how they look or how their voices sound* by something they are.
Automated teller machines (ATM) recognize customers from their
presentation of a bank card* something they have* and their entering a
personal identification number (PIN)* something they know. Using keys to
enter a locked building is another example of using something you have.
More secure systems may combine two or more of these approaches.
Generally, identifying travelers at the borders is performed by inspecting
their travel documents, such as passports and visas, and asking them
questions* things the travelers have and know. The U. S. Department of
State issues passports to U. S. citizens and visas to others who are not
U. S. citizens. The Immigration and Naturalization Service (INS) inspects
these travel documents at officially designated air, land, and sea ports
of entry.
Technologies called biometrics can automate the identification of
individual travelers by one or more of their distinct physical or
behavioral characteristics. Biometrics have been suggested as a way of
improving the nation*s ability to positively determine whether people are
admissible to the United States. The term biometrics covers a wide range
of technologies that can be used to verify identity by measuring and
Chapter 1: Introduction
Chapter 1: Introduction Page 19 GAO- 03- 174 Biometrics for Border
Security
analyzing human characteristics* relying on attributes of the individual
instead of things the individual may have or know. 1
Identifiable physiological characteristics include fingerprints, irises
and retinas, hand geometry, and facial geometry. How a person signs his or
her name is an example of an identifiable behavioral characteristic while
speech combines both physiological and behavioral characteristics. To be
effective identifiers, biometrics should be universally present, unique to
the individual, and stable over time. Biometrics theoretically represent a
more effective approach to security because each person*s biometric
characteristics are distinct and, when compared with identification cards
and passwords, are less easily lost, stolen, or guessed.
Biometrics have already been implemented to a limited degree in U. S.
border control systems. For example, the INS Passenger Accelerated Service
System (INSPASS) has identified travelers and expedited their inspections
at nine North American airports for almost 10 years. The Congress has
mandated a more extensive use of biometrics in automated border control
systems. A series of laws enacted between 1996 and spring 2002 requires
the federal government to develop Chimera, an automated information
system, to gather and share information among agencies about aliens
seeking to enter or stay in the United States. 2 The major requirements
for the Chimera system are (1) biometric identifiers; (2) machine-
readable visas, passports, and other travel and entry documents; and (3)
interoperability among all State Department, INS, and federal law
enforcement and intelligence agency systems that contain information about
aliens. Chimera will be used to screen applicants for visas and admission
to the United States, identify inadmissible and deportable aliens, track
lost and stolen passports, monitor foreign students studying
1 The term biometrics is commonly used to mean biometric technologies and
the characteristics themselves. 2 See 8 U. S. C. S:1365a and S:1722. These
laws* requirements reflect provisions of the Illegal Immigration Reform
and Immigrant Responsibility Act of 1996 (Public Law No. 104- 208, div. C,
S:110, Sept. 30, 1996), the INS Data Management Improvement Act of 2000
(Public Law No. 106- 215, June 15, 2000), the Uniting and Strengthening
America by Providing Appropriate Tools Required to Intercept and Obstruct
Terrorism Act of 2001 (USA PATRIOT Act) (Public Law No. 107- 56, S:403( c)
and S:414, Oct. 26, 2001), and the Enhanced Border Security and Visa Entry
Reform Act of 2002 (Public Law No. 107- 173, May 14, 2002). The Federally
Mandated Biometric Chimera System
Chapter 1: Introduction Page 20 GAO- 03- 174 Biometrics for Border
Security
in the United States, and help administer law enforcement and national
security. 3
The State Department, the Justice Department, and the National Institute
of Standards and Technology (NIST) were to report jointly to the Congress
by November 10, 2002, to assess the action needed to implement
machinereadable, tamper- resistant travel and entry documents and the
biometric comparison and authentication of such documents. By October 26,
2004, State and Justice are to issue to aliens only machine- readable,
tamperresistant visas and other travel and entry documents that use
biometric identifiers. At the same time, Justice is to install at all
ports of entry equipment and software that allow the biometric comparison
and authentication of all U. S. visas and other travel and entry documents
issued to aliens and machine- readable passports.
To provide the technological basis for Chimera by January 26, 2003, as
well as its supporting systems and databases, NIST is to develop a
technology standard, including biometric identifier standards for
verifying individual identities.
To address concerns about how information in the system will be used,
particularly with regard to privacy protection and security, the law
mandates that several steps be taken by October 26, 2002. First, the plan
for sharing law enforcement and intelligence information with the State
Department and INS must establish conditions for State*s and INS*s use of
the information that include their
limiting its redissemination;
ensuring that it is used solely for authorized purposes, with criminal
penalties for its misuse;
ensuring its accuracy, security, and confidentiality; 3 The information
in Chimera is to be accessible to federal law enforcement and intelligence
officers who, under federal regulation, are responsible for investigating
or identifying aliens (Enhanced Border Security and Visa Entry Reform Act,
S:202( a)( 5) (8 U. S. C. S:1722)), to federal law enforcement officials
to identify and detain individuals who pose a threat to national security
(USA PATRIOT Act, S:414( b) (8 U. S. C. S:1365a note)), and, at the
discretion of the attorney general, to federal, state, and local law
enforcement officials for law enforcement purposes (INS Data Management
Improvement Act, S:2 (8 U. S. C. S:1365a( f)( 2)), amending the Illegal
Immigration Reform and Immigrant Responsibility Act of 1996, S:110).
Chapter 1: Introduction Page 21 GAO- 03- 174 Biometrics for Border
Security
protecting privacy rights;
providing data integrity by removing obsolete and incorrect information;
and
protecting intelligence sources and methods. 4 Second, the Department of
State and the Department of Justice are to report jointly on the
*development, implementation, efficacy, and privacy implications* of a
*cross- agency, cross- platform electronic system* for sharing law
enforcement and intelligence information regarding aliens seeking to enter
the United States. 5
Third, the president is to establish a commission on interoperable data
sharing to oversee Chimera. 6 The commission*s duties include monitoring
the protections outlined above and considering recommendations regarding
security innovations, the adequacy of privacy protections, the adequacy of
mechanisms for correcting errors, and other protections against the
unauthorized use of data in the system.
This technology assessment focuses on four key questions: 1. What
biometric technologies are currently deployed, currently
available but not yet deployed, or in development that could be deployed
in the foreseeable future, for use in securing the nation*s borders?
2. How effective are these technologies now or likely to be in the future
in helping provide security to our borders?
3. What are the economic and effectiveness trade- offs of implementing
these technologies?
4 Enhanced Border Security and Visa Entry Reform Act, S:201( c)( 3) and
S:201( c)( 4) (8 U. S. C.
S:1356a note). The USA PATRIOT Act S:403( a) (amending 8 U. S. C. S:1105)
has virtually identical requirements with regard to the State Department*s
receiving National Crime Information Center data.
5 USA PATRIOT Act, S:403( c)( 2) and S:403( c)( 4). 6 Enhanced Border
Security and Visa Entry Reform Act, S:203 (8 U. S. C. S:1723).
An Overview of This Report
Chapter 1: Introduction Page 22 GAO- 03- 174 Biometrics for Border
Security
4. What are the implications of using biometric technologies for personal
security and the preservation of individual liberties?
To answer these questions, we first describe current border control
procedures for admitting people to the United States* issuing visas to
citizens of other nations and passports to U. S. citizens and inspecting
travelers at the ports of entry. Second, we describe how biometric
technologies work, including the different types of biometric
technologies, their levels of maturity, and their operating and
performance characteristics. We also describe current applications of
various biometric technologies.
We present four possible scenarios in which biometrics might be applied to
current U. S. border control procedures. For each scenario, we analyze
some of the costs, benefits, and risks associated with implementation.
Finally, we sum up the implications and challenges to be faced if a
biometric system is to be designed and deployed for border security.
Chapter 2: Today*s U. S. Border Control Procedures
Page 23 GAO- 03- 174 Biometrics for Border Security
Last year, there were more than half a billion border crossings into the
United States at almost 400 designated ports of entry. Many of these
border crossings were by travelers who crossed the border many times in 1
year, some daily. Table 4 shows that the vast majority of inspections*
those at border crossings* are at land ports. At land ports of entry in
fiscal year 2001, more than 414 million border crossers entered the United
States as one of more than 56 million pedestrians or in one of more than
140 million vehicles.
Table 4: Number of Inspections at U. S. Ports of Entry, Fiscal Year 2001
Type of port Number of ports Number of inspections
Sea 86 11,952,501 Air 155 79,598,681 Land 154 414,364,965
Total 395 505,916,147
Source: GAO analysis of INS data.
The laws and regulations governing entry into the United States and the
conditions of stay vary by citizenship and method of travel. 1 In general,
entry must be accompanied by the appropriate travel documents. U. S.
citizens generally must have a U. S. passport to leave or enter the United
States. Immigrants generally must have either a U. S. permanent resident
card or an immigrant visa and a passport from their own country.
Nonimmigrants generally must have a passport from their country and a
nonimmigrant visa. The numerous exceptions to these rules include the
following:
Passports are not required of U. S. citizens returning from Canada or
Mexico. 2
Passports are not required of Canadian citizens unless they are
returning from outside the Western Hemisphere. Visas are generally not
required for Canadian citizens.
1 The Immigration and Nationality Act of 1952, as amended (8 U. S. C.
S:1101 et seq.), and titles 8 and 22 of the Code of Federal Regulations
are the primary sources of U. S. immigration law.
2 According to the Department of Justice, passports are not required of U.
S. citizens returning from any point within the Western Hemisphere except
Cuba. Chapter 2: Today*s U. S. Border Control
Procedures
Chapter 2: Today*s U. S. Border Control Procedures
Page 24 GAO- 03- 174 Biometrics for Border Security
Passports and visas are not required of Mexican citizens who possess a
border crossing card issued by the U. S. government allowing them to enter
for business or pleasure.
Visas are not required of citizens of countries participating in the
visa waiver program who enter for business or pleasure. 3
The United States relies on two primary procedures to facilitate the entry
of people authorized to enter the country and to ensure that inadmissible
people are prevented from entering. The State Department*s Bureau of
Consular Affairs issues international travel documents, including
passports to U. S citizens and visas to people who are not U. S. citizens.
INS inspects travelers entering the United States through official ports
of entry. In addition, INS*s Border Patrol is responsible for securing the
borders and apprehending travelers entering through other than official
ports of entry.
U. S. citizens can apply for a passport at more than 4,500 passport
acceptance offices (see figure 1). Few of these are State Department
offices; most are offices in facilities such as U. S. post offices or
state, county, township, and municipal government offices. All first- time
applicants for a passport must appear before a passport acceptance agent
at one of these offices.
3 The visa waiver program permits nationals from designated countries to
apply for admission to the United States for 90 days or less as
nonimmigrant visitors for business or pleasure without first obtaining a
U. S. nonimmigrant visa. The following countries participate: Andorra,
Australia, Austria, Belgium, Brunei, Denmark, Finland, France, Germany,
Iceland, Ireland, Italy, Japan, Liechtenstein, Luxembourg, Monaco, the
Netherlands, New Zealand, Norway, Portugal, San Marino, Singapore,
Slovenia, Spain, Sweden, Switzerland, the United Kingdom, and Uruguay (8
U. S. C. S:1187, 8 C. F. R. S:217.2).
How U. S. Passports Are Issued
Chapter 2: Today*s U. S. Border Control Procedures
Page 25 GAO- 03- 174 Biometrics for Border Security
Figure 1: The U. S. Passport Application Process
Source: GAO adaptation of State Department data. Passport
Passport To renew
2 photos
Application
Old passport -OR
Central application processing
center Electronically keyed into U. S.
State Department*s
computer system
Name check
Multiple passports
Multiple applications
Passport examiner 16 passport centers
Decides Mail
Passport issued? or Send
results To apply
2 photos
Application
Proof of citizenship
Proof of identity More than 4,500 passport offices
Application package
Applicationpackage
Computer checks
Passport acceptance agent
Chapter 2: Today*s U. S. Border Control Procedures
Page 26 GAO- 03- 174 Biometrics for Border Security
Passport applicants must submit a passport application, proof of U. S.
citizenship, proof of identity, two passport photographs, and the
application fee. Passport acceptance agents, trained by the State
Department to look for potential fraud, review application packages and
may ask for additional documentation at their discretion. The agents fill
out an observation checklist that includes any concerns they have about
the validity of an applicant*s identity or citizenship documents. Passport
acceptance agents also are to ensure that the photographs match the
applicant. The acceptance agents send the application packages to a
central application processing center.
Applicants submitting renewal applications may mail them directly to the
central application processing center. The old passport, which can serve
as proof of identity and citizenship, is sent with the renewal
application. About 25 percent of the passport applications the State
Department receives arrive through the mail.
At the central application processing center, the application information
is electronically keyed into State*s computer system, and the application
package is forwarded to 1 of 16 State Department passport centers. State*s
computer systems conduct the following checks:
A name check, using the Consular Lookout and Support System (CLASS).
CLASS, which is used also before U. S. visas are issued, contains lookout
records of people who may be ineligible to receive a passport and is
populated from a variety of sources, including intelligence, immigration,
and child support enforcement data. CLASS also includes information on
passports and visas reported lost and stolen. Passport applicants are
checked against about 3.2 million records in CLASS.
A check to determine if the applicant already has an active U. S.
passport. An estimated 55 million U. S. passports are currently valid.
A check to determine if the applicant has multiple passport applications
in progress.
At the passport centers, passport examiners review each application,
including the results of the computer checks, and determine whether to
issue passports. A passport may be refused to an applicant for a variety
of reasons: The applicant may be subject to an outstanding federal warrant
Chapter 2: Today*s U. S. Border Control Procedures
Page 27 GAO- 03- 174 Biometrics for Border Security
for a felony, subject to a court order committing the applicant to a
mental institution, or in arrears for child support payments in excess of
$5,000. 4
A passport examiner looks at an entire application as a whole. A *hit* on
one of the computer checks does not necessarily result in a rejected
application. For example, some government officials who apply may have
both a personal passport and an official passport. The passport examiner
may resolve name check hits with other data such as place of residence or
Social Security number to differentiate between people who may have the
same name but are not the same person. If the examiner suspects a problem
with the application package, the case can be given to a fraud program
manager, who can perform a more detailed investigation, such as verifying
the authenticity of the identification or citizenship documents.
If the passport examiner is satisfied that the applicant*s documents are
authentic and that there is no reason to deny a passport, then the
examiner approves the application and the applicant is issued a U. S.
passport. Normally, the process takes about 6 weeks. Annually, the State
Department issues about 7 million passports that are valid for either 5 or
10 years, depending on the type of passport and the age of the applicant.
U. S. passports are depicted in figures 2 and 3.
4 Passports may be denied for reasons set forth in 22 C. F. R. S:51. 70.
Chapter 2: Today*s U. S. Border Control Procedures
Page 28 GAO- 03- 174 Biometrics for Border Security
Figure 2: A U. S. Passport Cover
Source: State Department.
Chapter 2: Today*s U. S. Border Control Procedures
Page 29 GAO- 03- 174 Biometrics for Border Security
Figure 3: A U. S. Passport*s Biography Page
Source: State Department.
With some exceptions, foreign visitors must present a visa to enter the
United States. Applicants can apply in person for an immigrant or
nonimmigrant visa at 210 American embassies or consulates (see figure 4).
The vast majority of issued U. S. visas are nonimmigrant visas. An
applicant for a nonimmigrant visa must submit an application, passport,
and photograph. 5 Some applications may be submitted by mail or in a drop
box outside the embassy or consulate. About 37 percent are submitted this
way.
5 The process for issuing immigrant visas, although similar to that for
nonimmigrant visas, includes other procedures and checks such as the
submission of an immigration petition to INS. About 628,000 immigrant
visas are issued each year. How U. S. Visas Are
Issued
Chapter 2: Today*s U. S. Border Control Procedures
Page 30 GAO- 03- 174 Biometrics for Border Security
Figure 4: The U. S. Visa Application Process
Source: GAO adaptation of State Department data. Submits Application
1 photo
Passport 210 American embassies and consulates
Mail or drop box
Name check
Multiple visas Possible
security advisory opinion required
State HQ Visa issued? Option
Decides Visa
or Visa Interview
by consular officer
Consular officer Application
package
Computer checks
Chapter 2: Today*s U. S. Border Control Procedures
Page 31 GAO- 03- 174 Biometrics for Border Security
After the data are keyed into the State Department*s visa computer system,
a consular officer reviews the application package. The officer may
interview the applicant, depending on the consular post and the type of
visa being applied for. Computer checks are conducted:
A name check, using CLASS, looks for any matches with individuals who
may be ineligible to receive a visa. Visa applications are checked against
about 6.5 million records in CLASS. 6 CLASS also includes records of lost
and stolen passports reported by other countries.
A check, using the Consular Consolidated Database (CCD), determines
whether the applicant has previously applied for a visa or currently has a
valid U. S. visa. CCD stores information about visa applications,
issuances, and refusals and obtains information about visa cases every 5
to 10 minutes from each consular post. CCD has about 58 million visa
records.
The consular officer makes a decision on whether to issue a visa, based on
information gathered from the visa application, passport, supporting
documentation, interview (if applicable), and computer checks. In some
cases, such as a name- check hit in CLASS, a security advisory opinion
from State Department headquarters may also be required. Visas may be
denied for a variety of reasons, including health- related reasons,
certain criminal offenses, and immigration violations. 7 Using fraudulent
documents to obtain a U. S. visa is also grounds for denial.
For nonimmigrant visas, the consular officer must be satisfied that an
applicant is not intending to become an immigrant. If the consular officer
is satisfied that the applicant*s documents are authentic and that there
is no reason to deny a visa, then the officer approves the application and
a visa is issued (see figure 5). The process can take from a day to
several weeks to complete. Last year, of the 10.5 million applications
received, about 7.5 million nonimmigrant visas were issued. Depending on
the type of visa and the nationality of the applicant, visas can be issued
for up to 10 years.
6 The State Department is adding 8 million criminal history alien records
from the Federal Bureau of Investigation (FBI). These records include
foreign- born individuals and individuals with unknown place of birth.
7 Visas may be denied for reasons listed in the Immigration and
Nationality Act, S:212 (8 U. S. C. S:1182).
Chapter 2: Today*s U. S. Border Control Procedures
Page 32 GAO- 03- 174 Biometrics for Border Security
Figure 5: A U. S. Visa Foil
Source: State Department.
Lawful entry into the United States generally must be completed through an
official air, land, or sea port of entry. Nearly 82 percent of the more
than 500 million inspections occur at land ports. Travelers* nationalities
and how they enter dictate the primary inspection procedures (see figure
6). A primary inspector is to question each traveler regarding his or her
identity and purpose for entering the United States. In addition, the
nspector can inspect a traveler*s travel documents and perform computer
checks on the traveler*s name or motor vehicle license plate. While the
State Department is responsible for initially granting or denying
permission to come to the United States, inspectors ultimately decide
whether to allow the traveler into the country at the ports of entry. The
issuance of a U. S. visa does not guarantee permission to enter.
Inspection at U. S.
Ports of Entry
Chapter 2: Today*s U. S. Border Control Procedures
Page 33 GAO- 03- 174 Biometrics for Border Security
Figure 6: The U. S. Port of Entry Inspection Process
Source: GAO adaptation of INS data.
At primary inspection, the INS inspector either permits travelers to enter
or refers them to secondary inspection, where a more detailed review of
the travel documents or further questioning can be conducted by another
INS inspector. People may be refused entry for the same reasons they can
be denied a visa. For U. S. citizens, once an inspector is convinced that
a traveler is a citizen, the inspection is considered complete for
immigration
Primary inspection
Document inspection
Name check Vehicle license plate check (land only)
Interview
If problem - secondary inspection
Thorough interviews
Comprehensive computer checks
J. DOE ? AOK- 11
If no problem Does admit If problem
Does not admit
Chapter 2: Today*s U. S. Border Control Procedures
Page 34 GAO- 03- 174 Biometrics for Border Security
purposes. However, checks can still be conducted to determine whether the
person is wanted by law enforcement authorities.
Overall, in fiscal year 2001, about 1.7 percent of travelers entering the
United States were referred to secondary inspection. Of those referred,
about 8 percent were denied admission to the United States. The numbers in
fiscal year 2001 were
primary inspections: 505, 916,147,
secondary inspections: 8,838,624, and
travelers denied admission: 707,920. At air ports of entry, commercial
carriers are required to submit passenger and crew manifests to INS
through the Advance Passenger Information System (APIS) for flights into
the United States. For each passenger, the first and last name, date of
birth, nationality, and passport number are transmitted. With information
from APIS, INS passenger analysis units can analyze intelligence on
passengers before flights arrive and identify passengers who will require
referral to secondary inspection.
Primary inspectors are to examine travel documents from all travelers at
air ports of entry. A name check is also to be conducted on all travelers,
using the Interagency Border Inspection System (IBIS). Machine- readable
passports are read with IBIS; the primary inspector manually types in the
names of travelers who do not have machine- readable passports. IBIS is a
multiagency database of lookout information that alerts inspectors of
conditions that may make travelers inadmissible to the United States. It
also provides information about warrants for U. S. citizens who may be
wanted by U. S. law enforcement agencies. IBIS contains data from law
enforcement and other agencies with inspection responsibilities at the
ports of entry, including the Animal Plant Health Inspection Service, the
Drug Enforcement Administration, and the Federal Bureau of Investigation
(FBI).
At sea ports of entry, some commercial carriers submit passenger manifests
to INS through APIS before docking. 8 As at airports, INS*s
8 In January 2003, INS plans to publish regulations in response to the
Enhanced Border Security and Visa Entry Reform Act to mandate electronic
manifest transmission from carriers at air and sea ports of entry for all
arriving and departing passengers.
Chapter 2: Today*s U. S. Border Control Procedures
Page 35 GAO- 03- 174 Biometrics for Border Security
passenger analysis units identify passengers who require further
examination when they enter the United States. At sea ports of entry
equipped with IBIS, the operation is very similar to that at an airport.
However, at most sea ports of entry, inspections are conducted aboard a
vessel. When the vessel docks, it is sealed so that no goods or persons
can be offloaded until it has been inspected.
INS inspectors board ships with the Portable Automated Lookout System
(PALS) housed on a laptop computer. PALS contains lookout information but
does not have as many records as IBIS and is not updated as often. INS
inspectors use PALS to perform name checks and examine documents of all
aliens aboard a vessel. For U. S. citizens, only documents are checked.
The inspection process on some of the larger cruise ships can take up to 6
hours to complete.
At land ports of entry, the procedures differ for pedestrians and those in
vehicles. In addition, at land ports, INS shares primary inspection
responsibilities with the Customs Service of the Treasury Department. INS
and Customs inspectors are cross- designated to perform each other*s
primary inspection duties so that either inspector may conduct the primary
inspection, following both INS and Customs procedures. INS has established
procedures to examine travelers expeditiously at many land ports of entry
because of the large volume of traffic at land crossings. Figure 7 shows
vehicles waiting at a U. S. land port of entry.
Chapter 2: Today*s U. S. Border Control Procedures
Page 36 GAO- 03- 174 Biometrics for Border Security
Figure 7: Motor Vehicles Waiting for Inspection at the Paso del Norte Port
of Entry, El Paso, Texas
Source: GAO.
For pedestrians at land ports of entry, generally all travelers* documents
are to be checked. If IBIS is available, a traveler*s name is either
machineread from the machine- readable passport or manually keyed in by an
inspector. U. S. citizens are not required to have a passport when
entering at a land port. Usually, they need only make an oral declaration
of U. S. citizenship. Similarly, at land ports of entry, Canadians are not
required to have a passport. Mexicans who possess a border crossing card
are not required to present either a Mexican passport or a U. S. visa. 9
Approximately 5 million border crossing cards have been issued to Mexican
nationals.
For vehicles at land ports of entry, license plates of all vehicles are to
be checked through IBIS. Some ports are equipped with automated license
plate readers. At others, an inspector manually keys license plate
9 A Mexican border crossing card permits the holder to enter for business
or pleasure and stay in the United States for 72 hours or less, going no
farther than 25 miles from the border.
Chapter 2: Today*s U. S. Border Control Procedures
Page 37 GAO- 03- 174 Biometrics for Border Security
information into IBIS as vehicles approach the inspection booth. As with a
name check, IBIS contains lookout information that alerts inspectors of
conditions that may make the occupants of a vehicle inadmissible.
Documents and names of the vehicle*s occupants are checked randomly or
when an inspector suspects that something is wrong. Figure 8 shows a
driver being questioned at a land port of entry.
Figure 8: A Driver Being Questioned at a Port of Entry
Source: U. S. Customs Service.
At land borders, aliens who require additional documentation, such as an
Arrival/ Departure Record, are to be referred to secondary inspection and
queried through IBIS. This includes aliens in possession of a nonimmigrant
visa and those traveling under the visa waiver program.
Some land ports of entry have implemented a program called Secure
Electronic Network for Travelers Rapid Inspection (SENTRI) to expedite the
inspection of vehicles and their occupants. With SENTRI, border crossers
register their vehicles and up to eight occupants, who are checked against
the IBIS database. Vehicles are identified when approaching a SENTRI-
equipped port of entry, using a transponder installed on the vehicles.
Pictures taken of each potential vehicle occupant at registration are
presented to the primary inspector on a computer
Chapter 2: Today*s U. S. Border Control Procedures
Page 38 GAO- 03- 174 Biometrics for Border Security
screen in the inspection booth when a vehicle drives up. The inspector
visually compares the pictures against the people in the vehicle. SENTRI
has reduced the average inspection time for each vehicle to about 10
seconds from the earlier 30 to 40 seconds.
Similar to SENTRI, other vehicle ports of entry have implemented a program
called NEXUS that is run jointly by the United States and Canada. Instead
of issuing a transponder to a vehicle, a proximity card is issued to each
registered traveler that is detected as a vehicle approaches the
inspection booth of a NEXUS- equipped port of entry. Photographs of
travelers detected by their proximity cards are presented to the primary
inspector, who can then verify the identity of each vehicle*s occupants.
Regardless of the method of entry, secondary inspection gives inspectors
more time with travelers to determine their admissibility than primary
inspection. In deciding whether to admit a traveler, the inspector reviews
the traveler*s documents for accuracy and validity and checks INS*s and
other agencies* databases for any information that could affect the
traveler*s entry, including criminal history information from the FBI and
nonimmigrant visa issuance data from the State Department. A fingerprint
identification system is also available in secondary inspection to
determine whether INS has apprehended the person for immigration offenses
or whether other law enforcement agencies are looking for the person.
Chapter 3: Biometric Technologies for Personal Identification
Page 39 GAO- 03- 174 Biometrics for Border Security
In this chapter, we define biometrics and explain how they work, describe
leading and emerging biometrics, and briefly introduce a few of the most
common applications of biometric technologies. In considering how to apply
biometrics to border control, we summarize data related to accuracy, the
lack of applications- dependent evaluations, systems* susceptibility to
deception, the status of standards, and users* acceptance. After briefly
comparing performance data on the technologies now considered most viable
for U. S. border control* facial, fingerprint, and iris recognition and
hand geometry* we end the chapter with a short list of biometric systems
in border control situations today, here and in other countries.
When used for personal identification, biometric technologies measure and
analyze human physiological and behavioral characteristics. Identifying a
person*s physiological characteristics is based on direct measurement of a
part of the body* fingertips, hand geometry, facial geometry, and eye
retinas and irises. The corresponding biometric technologies are
fingerprint recognition, hand geometry, and facial, retina, and iris
recognition. Identifying behavioral characteristics is based on data
derived from actions, such as speech and signature, the corresponding
biometrics being speaker recognition and signature recognition.
Biometrics are theoretically very effective personal identifiers because
the characteristics they measure are thought to be distinct to each
person. Unlike conventional identification methods that use something you
have, such as an identification card to gain access to a building, or
something you know, such as a password to log on to a computer system,
these characteristics are integral to something you are. Because they are
tightly bound to an individual, they are more reliable, cannot be
forgotten, and are less easily lost, stolen, or guessed.
Biometric technologies vary in complexity, capabilities, and performance,
but all share several elements. Biometric identification systems are
essentially pattern recognition systems. They use acquisition devices such
as cameras and scanning devices to capture images, recordings, or
measurements of an individual*s characteristics and computer hardware and
software to extract, encode, store, and compare these characteristics.
Because the process is automated, biometric decision making is generally
very fast, in most cases taking only a few seconds in real time. Chapter
3: Biometric Technologies for
Personal Identification Biometrics Defined How the Technologies Work
Chapter 3: Biometric Technologies for Personal Identification
Page 40 GAO- 03- 174 Biometrics for Border Security
Depending on the application, biometric systems can be used in one of two
modes: verification or identification. Verification* also called
authentication* is used to verify a person*s identity* that is, to
authenticate that individuals are who they say they are. Identification is
used to establish a person*s identity* that is, to determine who a person
is. Although biometric technologies measure different characteristics in
substantially different ways, all biometric systems involve similar
processes that can be divided into two distinct stages: enrollment and
verification or identification.
In enrollment, a biometric system is trained to identify a specific
person. The person first provides an identifier, such as an identity card.
The biometric is linked to the identity specified on the identification
document. He or she then presents the biometric (e. g., fingertips, hand,
or iris) to an acquisition device. The distinctive features are located;
one or more samples are extracted, encoded, and stored as a reference
template for future comparisons. Depending on the technology, the
biometric sample may be collected as an image, a recording, or a record of
related dynamic measurements. How biometric systems extract features and
encode and store information in the template are based on the system
vendor*s proprietary algorithms.
Template size also varies, depending on the vendor and the technology.
Although templates can range from 9 to 20,000 bytes, most are smaller than
1,000 bytes. Such small sizes allow for rapid comparison. Templates can be
stored remotely in a central database or within a biometric reader device
itself; their small size also allows for storage on smart cards or tokens.
Minute changes in positioning, distance, pressure, environment, and other
factors influence the generation of a template, making each template
likely to be unique, each time an individual*s biometric data are captured
and a new template is generated. Consequently, depending on the biometric
system, a person may need to present biometric data several times in order
to enroll. Either the reference template may then represent an amalgam of
the captured data or several enrollment templates may be stored. The
quality of the template or templates is critical in the overall success of
the biometric application. Because biometric features can change over
time, people may have to reenroll to update their reference template. Some
technologies can update the reference template during matching operations.
Enrollment
Chapter 3: Biometric Technologies for Personal Identification
Page 41 GAO- 03- 174 Biometrics for Border Security
The enrollment process also depends on the quality of the identifier the
enrollee presents. The reference template is linked to the identity
specified on the identification document. If the identification document
does not specify the individual*s true identity, the reference template
will be linked to a false identity.
In verification systems, the step after enrollment is to verify that a
person is who he or she claims to be (i. e., the person who enrolled).
After the individual provides whatever identifier he or she enrolled with,
the biometric is presented, which the biometric system captures,
generating a trial template that is based on the vendor*s algorithm. The
system then compares the trial biometric template with this person*s
reference template, which was stored in the system during enrollment, to
determine whether the individual*s trial and stored templates match (see
figure 9). Verification
Chapter 3: Biometric Technologies for Personal Identification
Page 42 GAO- 03- 174 Biometrics for Border Security
Figure 9: The Biometric Verification Process
Source: GAO.
Verification is often referred to as 1: 1 (one- to- one) matching.
Verification systems can contain databases ranging from dozens to millions
of enrolled templates but are always predicated on matching an
individual*s presented biometric against his or her reference template.
Nearly all verification systems can render a match* no- match decision in
less than a second. A system that requires employees to authenticate their
claimed identities
Stored Database Processed
Processed 10101110....
Reference template 10101110.... Trial template Enrollment
Multiple samples Verification 1: 1
Sample Yes No (Am I who I claim to be?)
Does trial template match reference template?
Reference template 10101110....
Chapter 3: Biometric Technologies for Personal Identification
Page 43 GAO- 03- 174 Biometrics for Border Security
before granting them access to secure buildings or to computers is a
verification application.
In identification systems, the step after enrollment is to identify who
the person is. Unlike verification systems, no identifier need be
provided. To find a match, instead of locating and comparing the person*s
reference template against his or her presented biometric, the trial
template is compared against the stored reference templates of all
individuals enrolled in the system (see figure 10). Identification systems
are referred to as 1: N (one- to- N, or one- to- many) matching because an
individual*s biometric is compared against multiple biometric templates in
the system*s database. Identification
Chapter 3: Biometric Technologies for Personal Identification
Page 44 GAO- 03- 174 Biometrics for Border Security
Figure 10: The Biometric Identification Process
Source: GAO.
There are two types of identification systems: positive and negative.
Positive identification systems are designed to ensure that an
individual*s biometric is enrolled in the database. The anticipated result
of a search is a match. A typical positive identification system controls
access to a secure building or secure computers by checking anyone who
seeks access against a database of enrolled employees. The goal is to
determine whether a person seeking access can be identified as having been
enrolled in the system.
Stored N reference templates
Database Processed
Processed 10101110....
Reference template (1) 10101110...
(2) 01010001... (3) 00101010...
(N) 01101001... 10101110.... Trial template Enrollment
Multiple samples Identification 1: N
Sample Yes No (Do you know who I am?) Does trial template match
1 of N reference templates? ......................
Chapter 3: Biometric Technologies for Personal Identification
Page 45 GAO- 03- 174 Biometrics for Border Security
Negative identification systems are designed to ensure that a person*s
biometric information is not present in a database. The anticipated result
of a search is a nonmatch. Comparing a person*s biometric information
against a database of all who are registered in a public benefits program,
for example, can ensure that this person is not *double dipping* by using
fraudulent documentation to register under multiple identities.
Another type of negative identification system is a surveillance system
that uses a watch list. Such systems are designed to identify people on
the watch list and alert authorities for appropriate action. For all other
people, the system is to check that they are not on the watch list and
allow them normal passage. The people whose biometrics are in the database
in these systems may not have provided them voluntarily. For instance, for
a surveillance system, the biometrics may be faces captured from mug shots
provided by a law enforcement agency.
No match is ever perfect in either a verification or an identification
system, because every time a biometric is captured, the template is likely
to be unique. Therefore, biometric systems can be configured to make a
match or no- match decision, based on a predefined number, referred to as
a threshold, that establishes the acceptable degree of similarity between
the trial template and the enrolled reference template. After the
comparison, a score representing the degree of similarity is generated,
and this score is compared to the threshold to make a match or no- match
decision. For algorithms for which the similarity between two templates is
calculated, a score exceeding the threshold is considered a match. For
algorithms for which the difference between two templates is calculated, a
score below the threshold is considered a match. Depending on the setting
of the threshold in identification systems, sometimes several reference
templates can be considered matches to the trial template, with the better
scores corresponding to better matches.
A growing number of biometric technologies have been proposed over the
past several years, but only in the past 5 years have the leading ones
become more widely deployed. Some technologies are better suited to
specific applications than others, and some are more acceptable to users.
Table 5 lists the seven leading biometric technologies we describe in this
section. Leading Biometric
Technologies
Chapter 3: Biometric Technologies for Personal Identification
Page 46 GAO- 03- 174 Biometrics for Border Security
Table 5: Leading Biometric Technologies and Their Template Size Technology
How it works Template size in bytes
Facial recognition Captures and compares facial patterns 84 or 1,300 a
Fingerprint recognition Captures and compares fingertip patterns 250*
1,000 Hand geometry Measures and compares dimensions of hand and fingers 9
Iris recognition Captures and compares iris patterns 512 Retina
recognition Captures and compares retina patterns 96 Signature recognition
Captures and compares rhythm, acceleration, and pressure flow of signature
1,000* 3,000 Speaker recognition Captures and compares cadence, pitch, and
tone of vocal tract 10,000* 20, 000
a Depending on the algorithm. Source: GAO analysis of manufacturer data.
Facial recognition technology identifies people by analyzing features of
the face not easily altered* the upper outlines of the eye sockets, the
areas around the cheekbones, and the sides of the mouth. The technology is
typically used to compare a live facial scan to a stored template, but it
can also be used in comparing static images such as digitized passport
photographs. Facial recognition can be used in both verification and
identification systems. In addition, because facial images can be captured
from video cameras, facial recognition is the only biometric that can be
used for surveillance purposes.
The two primary algorithms used in facial recognition systems are based on
the eigenface method and local feature analysis (LFA). The eigenface
method looks at the face as a whole and represents a person*s face as a
set of templates that require 1,300 bytes. LFA breaks down the face into
feature- specific fields, such as the eyes, nose, mouth, and cheeks,
creating an 84 byte template.
Fingerprint recognition is one of the best known and most widely used
biometric technologies. Automated systems have been commercially available
since the early 1970s, and there are currently more than 75 fingerprint
recognition technology companies. Until recently, it was used primarily in
law enforcement applications.
Fingerprint recognition technology extracts features from impressions made
by the distinct ridges on the fingertips. The fingerprints can be either
flat or rolled. A flat print captures only an impression of the central
area Facial Recognition
Fingerprint Recognition
Chapter 3: Biometric Technologies for Personal Identification
Page 47 GAO- 03- 174 Biometrics for Border Security
between the fingertip and the first knuckle; a rolled print captures
ridges on both sides of the finger.
An image of the fingerprint is captured by a scanner, enhanced, and
converted into a template. Scanner technologies can be optical, silicon,
or ultrasound technologies. Ultrasound, while potentially the most
accurate, has not been demonstrated in widespread use. Optical scanners
are the most commonly used. During enhancement, *noise* caused by such
things as dirt, cuts, scars, and creases or dry, wet, or worn fingerprints
is reduced, and the definition of the ridges is enhanced. Template size
ranges from 250 bytes up to 1,000 bytes, depending on which vendor*s
proprietary algorithm the system uses. Approximately 80 percent of vendors
base their algorithms on the extraction of minutiae points relating to
breaks in the ridges of the fingertips. Other algorithms are based on
extracting ridge patterns.
Hand geometry systems have been in use for almost 30 years for access
control to facilities ranging from nuclear power plants to day care
centers. Hand geometry technology measures the width, height, and length
of the fingers, distances between joints, and shapes of the knuckles.
Hand geometry systems use an optical camera and light- emitting diodes
with mirrors and reflectors to capture two orthogonal two- dimensional
images of the back and sides of the hand. Ninety- six measurements are
then extracted and a 9 byte template is derived, making it the smallest in
the biometric industry.
Although the basic shape of an individual*s hand remains relatively stable
over his or her lifetime, natural and environmental factors can cause
slight changes.
Iris recognition technology is based on the distinctly colored ring
surrounding the pupil of the eye. Made from elastic connective tissue, the
iris is a very rich source of biometric data, having approximately 266
distinctive characteristics. These include the trabecular meshwork, a
tissue that gives the appearance of dividing the iris radially, with
striations, rings, furrows, a corona, and freckles. Iris recognition
technology uses about 173 of these distinctive characteristics. Formed
during the eighth month of gestation, these characteristics reportedly
remain stable throughout a person*s lifetime, except in cases of injury.
Hand Geometry
Iris Recognition
Chapter 3: Biometric Technologies for Personal Identification
Page 48 GAO- 03- 174 Biometrics for Border Security
Iris recognition systems use a small, high- quality camera to capture a
black- and- white, high- resolution image of the iris. They then define
the boundaries of the iris, establish a coordinate system over the iris,
and define the zones for analysis within the coordinate system. The
visible characteristics within the zones are then converted into a 512
byte template that is used to identify or verify the identity of an
individual.
Retina recognition technology captures and analyzes the patterns of blood
vessels on the thin nerve on the back of the eyeball that processes light
entering through the pupil. Retinal patterns are highly distinctive
traits. Every eye has its own totally unique pattern of blood vessels;
even the eyes of identical twins are distinct. Although each pattern
normally remains stable over a person*s lifetime, it can be affected by
disease such as glaucoma, diabetes, high blood pressure, and autoimmune
deficiency syndrome.
The fact that the retina is small, internal, and difficult to measure
makes capturing its image more difficult than most biometric technologies.
An individual must position the eye very close to the lens of the retina-
scan device, gaze directly into the lens, and remain perfectly still while
focusing on a revolving light while a small camera scans the retina
through the pupil. Any movement can interfere with the process and can
require restarting. Enrollment can easily take more than a minute. The
generated template is only 96 bytes, one of the smallest of the biometric
technologies.
One of the most accurate and most reliable of the biometric technologies,
it is used for access control in government and military environments that
require very high security, such as nuclear weapons and research sites.
However, the great degree of effort and cooperation required of users has
made it one of the least deployed of all the biometric technologies.
Newer, faster, better retina recognition technologies are being developed.
Signature recognition authenticates identity by measuring handwritten
signatures. The signature is treated as a series of movements that contain
unique biometric data, such as personal rhythm, acceleration, and pressure
flow. Unlike electronic signature capture, which treats the signature as a
graphic image, signature recognition technology measures how the signature
is signed. Retina Recognition
Signature Recognition
Chapter 3: Biometric Technologies for Personal Identification
Page 49 GAO- 03- 174 Biometrics for Border Security
In a signature recognition system, a person signs his or her signature on
a digitized graphics tablet or personal digital assistant. The system
analyzes signature dynamics such as speed, relative speed, stroke order,
stroke count, and pressure. The technology can also track each person*s
natural signature fluctuations over time.
The signature dynamics information is encrypted and compressed into a
template that can range from slightly larger than 1,000 bytes to
approximately 3, 000 bytes. These templates are large by biometric
standards and reflect the variety of data available in a typical
signature.
Differences in how different people*s voices sound result from a
combination of physiological differences in the shape of vocal tracts and
learned speaking habits. Speaker recognition technology uses these
differences to discriminate between speakers.
During enrollment, speaker recognition systems capture samples of a
person*s speech by having him or her speak some predetermined information
into a microphone or telephone a number of times. This information, known
as a passphrase, can be a piece of information such as a name, birth
month, birth city, or favorite color or a sequence of numbers. Text
independent systems are also available that recognize a speaker without
using a predefined phrase.
This phrase is converted from analog to digital format, and the
distinctive vocal characteristics, such as pitch, cadence, and tone, are
extracted, and a speaker model is established. A template is then
generated and stored for future comparisons. Voice templates are much
larger than templates generated from other biometric technologies,
typically 10,000 to 20,000 bytes.
Speaker recognition can be used to verify a person*s claimed identity or
to identify a particular person. It is often where voice is the only
available biometric identifier, such as telephone and call centers.
Newer biometric technologies using diverse physiological and behavioral
characteristics are in various stages of development. Some are
commercially available, some may emerge over the next 2 to 4 years, and
others are many years from implementation. Table 6 lists the 9 we describe
in this section and their current maturity. Each technique*s Speaker
Recognition
Emerging Biometric Technologies
Chapter 3: Biometric Technologies for Personal Identification
Page 50 GAO- 03- 174 Biometrics for Border Security
performance can vary widely, depending on how it is used and its
environment in which it is used.
Table 6: Emerging Biometric Technologies and Their Maturity Technology How
it works Maturity
Vein scan Captures images of blood vessel patterns. Commercially
available. Facial thermography Infrared camera detects heat patterns
created by the branching
of blood vessels and emitted from the skin. Initial commercialization
attempts failed because of high cost.
DNA matching Compares actual samples of DNA rather than templates
generated from samples. Many years from implementation. Odor sensing
Captures the volatile chemicals that the skin*s pores emit. Years away
from commercial release. Blood pulse measurement Infrared sensors measure
blood pulse on a finger. Experimental. Skin pattern recognition Extracts
distinct optical patterns by spectroscopic measurement
of light scattered by the skin. Emerging. Nailbed identification An
interferometer detects phase changes in back- scattered light
shone on the fingernail; reconstructs distinct dimensions of the nailbed
and generates a one- dimensional map.
Emerging. Gait recognition Captures a sequence of images to derive and
analyze motion
characteristics. Emerging; requires further development.
Ear shape recognition Is based on distinctive ear shape and the structure
of the cartilaginous, projecting portion of the outer ear. Still a
research topic.
Source: GAO analysis.
Vein scan biometric technology can automatically identify a person from
the patterns of the blood vessels in the back of the hand. The technology
uses near- infrared light to detect vein vessel patterns. Vein patterns
are distinctive between twins and even between a person*s left and right
hand. Developed before birth, they are highly stable and robust, changing
throughout one*s life only in overall size. The technology is not
intrusive, and works even if the hand is not clean. It is commercially
available.
Facial thermography detects heat patterns created by the branching of
blood vessels and emitted from the skin. These patterns, called
thermograms, are highly distinctive. Even identical twins have different
thermograms. Developed in the mid- 1990s, thermography works much like
facial recognition, except that an infrared camera is used to capture the
images. The advantages of facial thermography over other biometric
technologies are that it is not intrusive* no physical contact is
required* every living person presents a usable image, and the image can
be collected on the fly. Also, unlike visible light systems, infrared
systems work accurately even in dim light or total darkness. Although
identification systems using facial thermograms were undertaken in 1997,
the effort was suspended because of the cost of manufacturing the system.
Chapter 3: Biometric Technologies for Personal Identification
Page 51 GAO- 03- 174 Biometrics for Border Security
DNA matching is a type of biometric in the sense that it uses a
physiological characteristic for personal identification. It is considered
to be the *ultimate* biometric technology in that it can produce
proofpositive identification of a person, except in the case of identical
twins. However, DNA differs from standard biometrics in several ways. It
compares actual samples rather than templates generated from samples.
Also, because not all stages of DNA comparison are automated, the
comparison cannot be made in real time. DNA*s use for identification is
currently limited to forensic applications. The technology is many years
away from any other kind of implementation and will be very intrusive.
Researchers are investigating a biometric technology that can distinguish
and measure body odor. This technology would use an odor- sensing
instrument (an electronic *nose*) to capture the volatile chemicals that
skin pores all over the body emit to make up a person*s smell. Although
distinguishing one person from another by odor may eventually be feasible,
the fact that personal habits such as the use of deodorants and perfumes,
diet, and medication influence human body odor renders the development of
this technology quite complex.
Blood pulse biometrics measure the blood pulse on a finger with infrared
sensors. This technology is still experimental and has a high false match
rate, making it impractical for personal identification.
The exact composition of all the skin elements is distinctive to each
person. For example, skin layers differ in thickness, the interfaces
between the layers have different undulations, pigmentation differs,
collagen fibers and other proteins differ in density, and the capillary
beds have distinct densities and locations beneath the skin. Skin pattern
recognition technology measures the characteristic spectrum of an
individual*s skin. A light sensor illuminates a small patch of skin with a
beam of visible and near- infrared light. The light is measured with a
spectroscope after being scattered by the skin. The measurements are
analyzed, and a distinct optical pattern can be extracted.
Nailbed identification technology is based on the distinct longitudinal,
tongue- in- groove spatial arrangement of the epidermal structure directly
beneath the fingernail. This structure is mimicked in the ridges on the
outer surface of the nail. When an interferometer is used to detect phase
changes in back- scattered light shone on the fingernail, the distinct
dimensions of the nailbed can be reconstructed and a one- dimensional map
can be generated.
Chapter 3: Biometric Technologies for Personal Identification
Page 52 GAO- 03- 174 Biometrics for Border Security
Gait recognition, recognizing individuals by their distinctive walk,
captures a sequence of images to derive and analyze motion
characteristics. A person*s gait can be hard to disguise because a
person*s musculature essentially limits the variation of motion, and
measuring it requires no contact with the person. However, gait can be
obscured or disguised if the individual, for example, is wearing loose
fitting clothes. Preliminary results have confirmed its potential, but
further development is necessary before its performance, limitations, and
advantages can be fully assessed.
Ear shape recognition is still a research topic. It is based on the
distinctive shape of each person*s ears and the structure of the largely
cartilaginous, projecting portion of the outer ear. Although ear
biometrics appears to be promising, no commercial systems are available.
Reduced cost, smaller size, greater accuracy, and greater ease of use are
making biometrics increasingly feasible for international travel
documentation, citizenship identification, automated banking, and benefits
dispersal. Biometrics have either been adopted or are being contemplated
for adoption in dozens of applications, ranging from modest* providing
time and attendance reports for small companies* to expansive* ensuring
the integrity of a registration database of 10 million voters.
Biometric systems have long been used to complement or replace badges and
keys in controlling access to entire facilities or specific areas within a
facility. The entrances to more than half the nuclear power plants in the
United States employ biometric hand geometry systems. They protected
athletes housed in Olympic Village at the 1996 games in Atlanta.
Recent reductions in the price of biometric hardware have spurred logical
access control applications. Fingerprint, iris, and speaker recognition
are replacing passwords to authenticate individuals accessing computers
and networks. The Office of Legislative Counsel of the U. S. House of
Representatives, for example, is installing an iris recognition system to
protect confidential files and working documents. Other federal agencies,
including the Department of Defense (DOD), Department of Energy, and
Department of Justice, as well as the intelligence community, are adopting
similar technologies. Common Applications
of Biometric Technologies
Access Control
Chapter 3: Biometric Technologies for Personal Identification
Page 53 GAO- 03- 174 Biometrics for Border Security
Leading banks and other financial service companies are experimenting with
facial, iris, and speaker recognition systems to authenticate ATM users
and to combat credit and debit card fraud. Hand geometry and iris and
facial recognition have been deployed at ATMs in North America, Europe,
and Asia. The JPMorgan Chase Bank allows some customers to access accounts
by speaker recognition. To address concerns about security and fraud,
organizations that offer Internet shopping are also considering biometric
technologies to authorize various types of transactions.
Biometrics can also be used in monitoring applications. Adding biometrics
to time and attendance processes, for example, helps prevent hourly
employees from punching time cards for their absent friends, a practice
that is estimated to cost employers hundreds of millions of dollars
annually. Biometrics are also being applied to prevent prison inmates from
swapping identities with visitors as they leave prisons.
In addition, biometric technologies are being used in large- scale
identification systems to determine whether applicants are already
enrolled under a different identity. One specific application has been to
prevent individuals from cheating public sector benefits programs by
collecting benefits under multiple identities. A number of states have
made fingerprinting a requirement for registration for welfare and other
types of public aid. Since biometric systems were deployed, the number of
individuals claiming benefits has dropped dramatically in several states
that use such systems. Internationally, in the Philippines, South Africa,
and Spain, programs to streamline or legitimize issuing government
benefits have enrolled millions of citizens.
Several states have implemented biometric systems to stop drivers,
particularly truck drivers, from maintaining duplicate licenses or
swapping licenses when crossing state lines or national borders. Large-
scale identification systems are also being used to register voters for
national and local elections to prevent voter fraud. Mexico, for example,
uses facial recognition technology to check voter rolls for duplicates in
its national elections. Brazil, Costa Rica, the Dominican Republic,
Panama, and Italy use fingerprints to verify voters at polling stations.
Criminal identification is far and away the oldest, most widespread,
largescale identification use of biometric systems. Automated fingerprint
recognition systems are employed around the world to identify suspects
Fraud Reduction
Licensing and Voter Applications
Criminal Identification and Surveillance
Chapter 3: Biometric Technologies for Personal Identification
Page 54 GAO- 03- 174 Biometrics for Border Security
within local, state, or federal databases of known offenders. Facial
recognition is also being used for criminal identification, although the
technology does not provide the same high degree of accuracy as the older
technology. Employee background checks are another application of large-
scale systems. The governments of Argentina, China, Nigeria, and Yemen are
all planning to implement biometrics in their national identification
programs.
Surveillance is one of the most recent applications of biometric systems.
Although the majority of the major casinos in North America have deployed
facial recognition surveillance systems for some time to spot known
cheaters, systems are now publicly deployed in Newham Borough, England;
Tampa, Florida; and Canada*s Lester B. Pearson International Airport in
Toronto. More recently, they have been used sporadically at such major
events as the 2001 Super Bowl in Tampa, Florida, and the winter Olympics
at Salt Lake City in 2002.
Biometric technologies are maturing but are still not widespread or
pervasive because of performance issues, including accuracy, the lack of
applications- dependent evaluations, their potential susceptibility to
deception, the lack of standards, and questions of users* acceptance.
These issues should be kept in mind when considering biometrics for U. S.
border control.
Biometrics is a very young technology, having only recently reached the
point at which basic matching performance can be acceptably deployed. It
is necessary to analyze several metrics to determine the strengths and
weaknesses of each technology and vendor for a given application.
The three key performance metrics are false match rate (FMR), false
nonmatch rate (FNMR), and failure to enroll rate (FTER). A false match
occurs when a system incorrectly matches an identity, and FMR is the
probability of individuals being wrongly matched. In verification and
positive identification systems, authorized people can be granted access
to facilities or resources as the result of incorrect matches. In a
negative identification system, the result of a false match may be to deny
access. For example, if a new applicant to a public benefits program is
falsely matched with a person previously enrolled in that program under
another identity, the applicant may be denied access to benefits. The FMR,
sometimes called the false positive rate, is sometimes confused with the
false accept rate. The FMR is the probability of an erroneous match in a
Performance Issues
Accuracy
Chapter 3: Biometric Technologies for Personal Identification
Page 55 GAO- 03- 174 Biometrics for Border Security
single template comparison while the false accept rate is a system measure
that a person is erroneously matched, combining the results of all
template comparisons. For example, in an identification match, the FMR
would be the probability that the trial template erroneously matches a
single selected reference template. The false accept rate would be the
probability that the trial template erroneously matches any of the
reference templates.
A false nonmatch occurs when a system rejects a valid identity, and FNMR
is the probability of valid individuals being wrongly not matched. In
verification and positive identification systems, people can be denied
access to some facility or resource as the result of a system*s failure to
make a correct match. In negative identification systems, the result of a
false nonmatch may be that a person is granted access to resources to
which she should be denied. For example, if a person who has enrolled in a
public benefits program under another identity is not correctly matched,
she will succeed in gaining fraudulent access to benefits. The FNMR,
sometimes called the false negative rate, is sometimes confused with the
false reject rate. The relationship between FNMR and the false reject rate
is similar to the relationship between the FMR and the false accept rate.
The FNMR is the probability of an erroneous nonmatch for a single template
comparison, while the false reject rate is a system measure that a person
is erroneously not matched, combining the results of all template
comparisons.
False matches may occur because there is a high degree of similarity
between two individuals* characteristics. False nonmatches occur because
there is not a sufficiently strong similarity between an individual*s
enrollment and trial templates, which could be caused by any number of
conditions. For example, an individual*s biometric data may have changed
as a result of aging or injury. If biometric systems were perfect, both
error rates would be zero. However, because biometric systems cannot
identify individuals with 100 percent accuracy, a trade- off exists
between the two.
False match and nonmatch rates are inversely related; they must therefore
always be assessed in tandem, and acceptable risk levels must be balanced
with the disadvantages of inconvenience. For example, in access control,
perfect security would require denying access to everyone. Conversely,
granting access to everyone would result in denying access to no one.
Obviously, neither extreme is reasonable, and biometric systems must
operate somewhere between the two.
Chapter 3: Biometric Technologies for Personal Identification
Page 56 GAO- 03- 174 Biometrics for Border Security
For most applications, how much risk one is willing to tolerate is the
overriding factor, which translates into determining the acceptable FMR.
The greater the risk entailed by a false match, the lower the tolerable
FMR. For example, an application that controlled access to a secure area
would require that the FMR be set low, which would result in a high FNMR.
However, an application that controlled access to a bank*s ATM might have
to sacrifice some degree of security and set a higher FMR (and hence a
lower FNMR) to avoid the risk of irritating legitimate customers by
wrongly rejecting them. This is displayed in figure 11.
Figure 11: The General Relationship between FMR and FNMR
Note: Equal error rate is the point at which FMR equals FNMR. Source: GAO.
1.0 False match rate (FMR) 0.75
0.5 0.25
0 0
False nonmatch rate (FNMR) 0.25 0.5 0.75 1.0 Equal error rate (EER)
Chapter 3: Biometric Technologies for Personal Identification
Page 57 GAO- 03- 174 Biometrics for Border Security
As figure 11 shows, selecting a lower FMR increases the FNMR. Perfect
security would require setting the FMR to 0, in which case the FNMR would
be 1. At the other extreme, setting the FNMR to 0 would result in an FMR
of 1.
The expectations regarding FMR and FNMR are very different for
verification and identification systems. In a verification system, a user
is checked against one or a few reference templates to confirm the user*s
claimed identity. A much higher standard is required for identification
systems where checks are made against all reference templates in the
database. Consequently, a much lower FMR is required for a large- scale
positive identification system than for a similar size verification
system, simply because even a small percentage of false matches for a
system that performed billions of comparisons a day would overwhelm the
resources dedicated to investigating positive matches. The larger the
identification database, the lower the false match rate needs to be to
maintain the number of false positives at a manageable amount.
Vendors often use equal error rate (EER), an additional metric derived
from FMR and FNMR, to describe the accuracy of their biometric systems.
EER refers to the point at which FMR equals FNMR (see figure 11). Setting
a system*s threshold at its EER will result in the probability that a
person is falsely matched equaling the probability that a person is
falsely not matched. However, this statistic tends to oversimplify the
balance between FMR and FNMR, because in few real- world applications is
the need for security identical to the need for convenience.
FTER is a biometric system*s third critical accuracy metric. FTER measures
the probability that a person will be unable to enroll. Failure to enroll
(FTE) may stem from an insufficiently distinctive biometric sample or from
a system design that makes it difficult to provide consistent biometric
data. The fingerprints of people who work extensively at manual labor are
often too worn to be captured. A high percentage of people are unable to
enroll in retina recognition systems because of the precision such systems
require. People who are mute cannot use voice systems, and people lacking
fingers or hands from congenital disease, surgery, or injury cannot use
fingerprint or hand geometry systems. Although between 1 and 3 percent of
the general public does not have the body part required for using any one
biometric system, they are normally not counted in a system*s FTER.
Chapter 3: Biometric Technologies for Personal Identification
Page 58 GAO- 03- 174 Biometrics for Border Security
Because biometric systems based solely on a single biometric may not
always meet performance requirements, the development of systems that
integrate two or more biometrics is emerging as a trend. Multiple
biometrics could be two types of biometrics, such as combining facial and
iris recognition. Multiple biometrics could also involve multiple
instances of a single biometric, such as 1, 2, or 10 fingerprints, 2
hands, and 2 eyes. One prototype system integrates fingerprint and facial
recognition technologies to improve identification. A commercially
available system combines face, lip movement, and speaker recognition to
control access to physical structures and small office computer networks.
Depending on the application, both systems can operate for either
verification or identification. Experimental results have demonstrated
that the identities established by systems that use more than one
biometric could be more reliable, be applied to large target populations,
and improve response time.
Biometric companies have primarily been concerned with testing the
accuracy of their technologies in highly controlled environments, using
static or artificially generated templates, images, and data. The results
of their tests, as quoted by vendors, are quite extraordinary, such as
claims of FMRs of 1 in 100,000, 1 in a billion, or even 1 in 10 78 and
FNMRs in the vicinity of 1 percent, 0.1 percent, and 0. 01 percent.
However, because the performance of a technology depends greatly on how
and where it is deployed, such numbers have proven to be far more
impressive than reallife performance data.
Until recently, there was no set methodology for testing the same
technologies in different applications. A recently developed methodology
uses a three- step evaluation protocol: a technology evaluation, followed
by a scenario evaluation and an operational evaluation of biometric
systems. 1 Each of the methodology*s three types of evaluation requires a
different protocol and produces different results. A technology evaluation
compares competing algorithms from a single technology to identify the
most promising approaches. A scenario evaluation tests overall system
performance for a class of applications under conditions that model
realworld applications. An operational evaluation measures performance for
a specific biometric system for a specific application in the actual
operating environment with actual users of the system. The Facial
Recognition
1 P. Jonathon Phillips and others, *An Introduction to Evaluating
Biometric Systems,* IEEE Computer 33: 2 (2000): 56* 63. Multimodal
Biometrics
The Lack of ApplicationsDependent Evaluations
Chapter 3: Biometric Technologies for Personal Identification
Page 59 GAO- 03- 174 Biometrics for Border Security
Vendor Test 2000 (FRVT 2000), which assessed the capabilities of
commercially available facial recognition systems, was based on this
evaluation methodology and included elements of technology and scenario
evaluations. 2
Studies by respected organizations in the United States and the United
Kingdom have provided a number of effective measures of the actual
performance of biometric systems in different real- world environments.
Sandia National Laboratories* 1996 evaluation of an iris recognition
identification system in an access- control environment included FNMRFMR
results. The International Biometric Group (IBG) has since 1999 conducted
side- by- side comparative performance testing of leading biometric
identity verification systems under real- world operating conditions. Test
results have included FNMRs, FMRs, and FTERs for fingerprint, iris,
facial, voice, keystroke, and signature systems. In 2000, the British
National Physical Laboratory (NPL) tested biometric identity verification
systems, including fingerprint, hand, iris, facial, voice, and vein, in
real- world environments. 3 FNMRs, FMRs, and FTERs were reported. The U.
S. Army Research Laboratory*s pilot study of iris and facial recognition
systems in 2000* 01 reported performance results that included error rates
as well as user perception and acceptability. Table 7 lists the
significant independent tests and their results since 1991.
2 FRVT 2000 was sponsored by the DOD Counterdrug Technology Development
Program Office, Defense Advanced Research Projects Agency, and National
Institute of Justice. The test goals were to know the strengths and
weaknesses of each individual system, understand the current state of the
art for facial recognition, and educate the community and general public
on how to present and analyze results.
3 NPL in the United Kingdom is analogous to the National Institute of
Standards and Technology (NIST) in the United States.
Chapter 3: Biometric Technologies for Personal Identification
Page 60 GAO- 03- 174 Biometrics for Border Security
Table 7: Independent Biometric Test Results, 1991* 2002 Test name Who
conducted Date Technology Type Performance measure
Test of Biometric Technologies Sandia National
Laboratories 1991 Fingerprint, hand, retina, signature,
speaker Technology FMR, FNMR, accept time
Hand Geometry Field Application Sandia National
Laboratories 1995 Hand Scenario Varied lighting, maintenance
IriScan Prototype Identifier Sandia National
Laboratories 1996 Iris Scenario FMR, FNMR, enrollment
time, transaction time Speaker Recognition Evaluations
National Institute of Standards and Technology
1996 to present Speaker Technology Handset variation, test
segment duration, speaker tracking, 1- speaker and 2- speaker, cellular
data Philippine AFIS Benchmark Test National Biometric
Test Center 1997 Fingerprint Technology FMR, FNMR SENTRI Test INS 1998
Facial, speaker Scenario FNMR Comparative Biometric Testing IBG 1999 to
present Facial, fingerprint, iris, keystroke,
signature, speaker Scenario FMR, FNMR, enrollment
rate, ergonomics, ease of use, temporal Biometric Product Testing NPL 2000
Facial, fingerprint,
hand, iris, vein, speaker
Scenario Failure to enroll and acquire, FMR, FNMR, transaction time, male
versus female FRVT 2000 DOD, National
Institute of Justice, NIST
2000 Facial Scenario, technology Probability of identity,
probability of verification, distance, temporal, expressions, pose,
resolution, media Fingerprint Verification Competition 2000
University of Bologna, Michigan State University, San Jose State
University
2000 Fingerprint Technology Enrollment time, matching time, EER
Facial Recognition Technology Department of State,
Bureau of Consular Affairs
2001 Facial Technology FNMR Personnel Identification Pilot Study
Army Research Laboratory 2001 Facial, iris Operational FMR, FNMR
Fingerprint Identification Device Federal Aviation
Administration and Safe Skies
2001 Fingerprint Operational FMR, FNMR, enrollment and transit time,
abnormal conditions (oil, grease, powder, injury, moist or dry skin,
offset angle, contact pressure, backlighting, attempts to defeat) Hand
Geometry Identification Device FAA and Safe Skies 2001 Hand Operational
FMR, FNMR, enrollment
and transit time, abnormal conditions (rings, injuries, backlighting,
attempts to defeat)
Chapter 3: Biometric Technologies for Personal Identification
Page 61 GAO- 03- 174 Biometrics for Border Security
Test name Who conducted Date Technology Type Performance measure
Facial Recognition Device FAA and Safe Skies 2002 Facial Operational FMR,
FNMR, enrollment
and transit time, abnormal conditions (glasses, facial hair, backlighting,
bandages, false photograph) Iris Recognition Device FAA and Safe Skies
2002 Iris Operational FTE, FNMR, enrollment and
transit times Biometric Security Test c*t Magazine 2002 Iris, fingerprint,
facial Technology Attempts to defeat Fingerprint Verification Competition
2002
University of Bologna, Michigan State University, San Jose State
University
2002 Fingerprint Technology Enrollment time, matching time, EER, FMR
Facial Recognition Vendor Test 2002 15 agencies and
organizations, including DOD, National Institute of Justice, and NIST
In progress Facial Scenario, technology In progress
Source: GAO analysis of independent biometric test results.
A rash of new tests of biometric systems has recently been initiated. The
results are likely to provide more sound means of evaluating the strengths
and weaknesses of the different technologies and vendors* products.
Can biometric systems be defeated? Many vendors claim that their systems
cannot be fooled because they are able to detect whether or not an
individual*s presented biometric is a live sample. Many biometric devices
can, in principle, determine whether a live characteristic is being
presented. Some fingerprint systems, for example, test for *liveness* by
relying on the unique conductive nature of live fingers. Others measure
blood flow or ensure that the ridges at the periphery of a print are
arrayed the same as in normal finger placement.
Although hand geometry systems do not actually check for a live biometric,
fingers have to be positioned so that they put pressure on the correct
pegs. Facial recognition checks for *liveness* by requiring users to
change their facial expression* by blinking their eyes or smiling, for
example* in order to successfully generate a template. With iris
recognition, light shone on the eye can be varied for recording pupil
dilation. Some speaker recognition systems can generate a random sequence
of numbers for each verification to ensure that a recorded voice is not
being played back. Moreover, low- fidelity recording devices are generally
not able to capture the high and low frequencies necessary for
verification. Susceptibility to Deception
Chapter 3: Biometric Technologies for Personal Identification
Page 62 GAO- 03- 174 Biometrics for Border Security
Nevertheless, recent tests are casting doubt on vendors* claims regarding
the maturity and security of their technologies. German technology
magazine c*t carried out tests on 11 commercially available biometric
systems used to control access to computers. 4 Facial, fingerprint, and
iris recognition systems were defeated by testers using photographs and
videos, reactivated latent images, and forgeries.
They spoofed one fingerprint recognition system by reactivating latent
fingerprints left on the surface of its capacitive sensor, simply by
breathing on the prints, placing a thin- walled water- filled plastic bag
on the sensor*s surface, and dusting the prints with graphite powder and
gently applying pressure to an adhesive film stretched over them. They
outfoxed another fingerprint recognition system whose optical scanner
required that an object be resting on its surface by creating a silicone
copy of a fingerprint of an enrolled person from a candle wax mold.
They spoofed an iris recognition system by using a high- resolution
printed picture of an enrolled person*s iris with a live person*s pupil
shining through a miniature hole cut out of the picture*s pupil. They beat
a wellknown facial recognition system by using a laptop computer to play
back
*live* images of an enrolled person to the camera. They fooled another
facial system by holding up a photograph of an enrolled person.
In another recent test, an engineering professor demonstrated how 11
commercially available fingerprint biometric systems could all be fooled
with a molded gelatin finger. A further recent test revealed that
biometric systems could be defeated by cracking the code of the templates
stored inside them. Using manufactured images that displayed the
characteristics required by the matching software, the tester defeated
commercially available fingerprint and retina recognition systems. These
tests certainly call into question the claim that biometric systems cannot
be deceived.
Identifying, exchanging, and integrating information from different and
perhaps unfamiliar sources and functions are essential to an effective
biometrics application. Without predefined standards, system developers
may need to define in detail the precise steps for exchanging information,
4 Lisa Thalheim, Jan Krissler, and Peter- Michael Ziegler, *Body Check:
Biometric Access Protection Devices and Their Programs Put to the Test,*
trans. Robert Smith, c*t Magazine
11 (2002): 114. The Development of
Biometric Standards
Chapter 3: Biometric Technologies for Personal Identification
Page 63 GAO- 03- 174 Biometrics for Border Security
a potentially complex, time- consuming, and expensive process. The risks
associated with not adopting standards for a system are significant,
because of the length of time the system must remain operational and the
rapid pace of technological change. The proprietary technology of choice
today may not be cost- effective or even supported tomorrow.
Attempts to standardize biometrics are under way in various areas, such as
the mechanics of image capture, the accuracy of data as they are
extracted, and device interoperability. However, the majority of biometric
devices and their software are still proprietary in many respects. For
example, the method for extracting features from a biometric sample such
as a fingerprint differs among most, if not all, vendors. Templates
containing biometric data, time stamps, encryption features, and device
information are also not standard. Devices from company A do not
necessarily work compatibly with devices from companies B and C.
Incompatibility is also an issue for communication between devices and
host computers, since programs are developed from vendors* software
development kits. Each vendor designs a software development kit for its
own products, so that the programs developed for one vendor*s product
generally cannot be used with another vendor*s products.
The biometrics community does employ several standards, however. We list
seven:
The wavelet scalar quantization (WSQ) gray- scale fingerprint image
compression algorithm is the standard for exchanging fingerprint images
within the criminal justice community. WSQ defines a class of encoders and
a single decoder with sufficient generality to decode compressed image
data produced by any compliant encoder.
The National Institute of Standards and Technology (NIST) issued the
Common Biometric Exchange File Format (CBEFF) on January 3, 2001. The
standard is designed to (1) facilitate biometric data interchange between
different system components or between systems, (2) promote the
interoperability of biometric- based application programs and systems, (3)
provide forward compatibility for technology improvements, and (4)
simplify the integration of software and hardware from different vendors.
BioAPI* Consortium has developed BioAPI, a specification for a highlevel
generic biometric authentication model suited for any form of biometric
technology. It covers the basic functions of enrollment, verification, and
identification and includes a database interface to
Chapter 3: Biometric Technologies for Personal Identification
Page 64 GAO- 03- 174 Biometrics for Border Security
allow a biometric service provider to manage the identification population
for optimum performance. It also provides methods that allow an
application to manage the capture of samples on a client and the
enrollment, verification, and identification on a server. While it does
not define security requirements for biometric applications and service
providers, it does explain how the application programming interface (API)
is intended to support good security practices.
In May 2000, Microsoft Corp. and I/ O Software Inc. announced that they
would cooperate to foster the widespread growth of biometrics through the
integration of biometric authentication technology in future versions of
the Microsoft Windows operating system. The resulting biometric
application programming interface (BAPI) is expected to define a standard
software protocol and API for communication between software applications
and biometric devices running on Microsoft Windows platforms. BAPI is
expected to standardize the way different biometric devices, such as
fingerprint scanners and facial recognition devices, communicate with the
application software that uses them. It is also expected to be a
comprehensively modular architecture that covers a variety of hardware
interfaces, encryption, biometric algorithms, and application interfaces.
Established by the Joint Photographic Experts Group (JPEG), the JPEG
specification can be used in facial recognition systems. 5 It describes an
image compression system that allows great flexibility not only for the
compression of images but also for access to the compressed data. The
specification is designed for compressing either full- color or gray-
scale images of natural, real- world scenes, although the decompressed
images are not quite the same as the originals. JPEG*s algorithm is
designed to exploit known limitations of the eye, notably that the eye
perceives small color changes less accurately than small changes in
brightness. This is a limitation if an application uses a JPEG image to
machine- analyze images, since the small errors JPEG introduces may be a
problem even if they are invisible to the eye. 6
5 JPEG members are experts nominated by national standards bodies and
major companies to produce standards for continuous tone image coding.
*Joint* refers to the group*s status as a committee working on standards
for both the International Organization for Standardization and
International Telecommunication Union* Telecommunication.
6 According to a February 2001 study conducted for the FBI, WSQ and JPEG
2000 formats are similar enough that questions may emerge about migration
of the FBI standard to the JPEG 2000 standard. Such questions would
include weighing some advantages against other disadvantages of changing
an accepted standard that is already widely used.
Chapter 3: Biometric Technologies for Personal Identification
Page 65 GAO- 03- 174 Biometrics for Border Security
In February 2001, the American National Standards Institute (ANSI)
approved the Biometric Information Management and Security (ANSI X9.84-
2001) standard. This standard specifies the minimum security requirements
for effective management of biometric data. The standard defines message
formats for carrying biometric data in a secure way and also defines many
concepts and procedures for the creation of a secure biometric system. The
message formats specified by X9.84 are more flexible than the BioAPI data
format because they allow a richer description of the biometric data and
are extensible. Moreover, the X9.84 standard addresses the issue of
integrity and privacy of biometric samples and templates in a flexible
way, by providing several different security mechanisms among which the
user can choose.
The American Association for Motor Vehicle Administration (AAMVA)
included a format for fingerprint minutiae data in its Driver License and
Identification (DL/ ID- 2000) Standard, which provides a uniform means to
identify issuers and holders of driver*s license cards within the United
States and Canada. The standard describes required and optional data
elements to be placed on a driver*s license card. Required elements
include the name, address, and photograph of the driver. While
fingerprints are classified as an optional data element, the standard
describes a way to record minutiae data based on the type, position,
angle, and quality of the minutiae point. A field is also provided for
recording vendor- specific data about the fingerprint. The biometric
portions of this standard are compatible with the BioAPI specification and
CBEFF.
Figure 12 shows the relationship of these standards to the individual
functional components necessary to make up a comprehensive biometric
system.
Chapter 3: Biometric Technologies for Personal Identification
Page 66 GAO- 03- 174 Biometrics for Border Security
Figure 12: Standards for Biometric Systems
Note: AAMVA = The American Association for Motor Vehicle Administration*s
Driver License and Identification (DL/ ID- 2000) Standard. WSQ = wavelet
scalar quantization. JPEG = a specification of the Joint Photographic
Experts Group. CBEFF = the National Institute of Standards and
Technology*s Common Biometric Exchange File Format. BioAPI = the BioAPI*
Consortium*s BioAPI specification for a high- level generic biometric
authentication model. BAPI = biometric application programming interface.
X9. 84 = the American National Standards Institute*s ANSI X9. 84- 2001
standard.
Source: GAO analysis of biometric standards.
Although a number of such standards have been developed, those required
for integrating all vendors* products are not yet available for all types
of applications. For example, the standard for how to store biometric
templates is not yet available. While the AAMVA standard describes a
common way to record fingerprint minutiae, it still allows for including
Hardware interface
BioAPI
BAPI Biometrics
capture device
Software interface
BioAPI
BAPI Data transport
CBEFF
BioAPI
BAPI
X9.84 Biometric
sample
WSQ
JPEG Biometric
template
AAMVA
Chapter 3: Biometric Technologies for Personal Identification
Page 67 GAO- 03- 174 Biometrics for Border Security
data in a vendor- specific format. Biometric templates, which capture only
the critical data needed to make a positive confirmation, are small and
can be stored on smart cards, but the template one vendor uses cannot
generally be used by another for some biometric technologies, such as
fingerprints. Working with other groups* the Biometric Consortium, the
BioAPI* Consortium, the Biometric Foundation, and the International
Biometric Industry Association (IBIA), among others* the InterNational
Committee for Information Technology Standards (INCITS) is reviewing draft
project proposals for standardizing biometric templates. 7 Without a
biometric template standard, it could be necessary to store the larger
biometric sample as well as the biometric template for each user during
enrollment. Such a standard would also allow for changes to the biometric
capture device (i. e., a change in equipment) or algorithms without
reenrolling all system users.
In November 2001, the executive board of INCITS established Technical
Committee M1, Biometrics, for the rapid development and approval of formal
national and international generic biometric standards. The goal of M1*s
work is to accelerate the deployment of significantly better, standards-
based security solutions for purposes such as homeland defense and the
prevention of identity theft, as well as other government and commercial
applications based on biometric personal authentication. INCITS approved
the BioAPI Specification, Version 1.1, as the ANSI/ INCITS 358- 2002*
Information technology* BioAPI Specification, on February 13, 2002. It is
now considering CBEFF for fast track processing in the near future.
Additionally, M1 is now reviewing contributions of draft project proposals
for the standardization of biometric templates. M1 is also anticipating
contributions of draft project proposals for the development of
application profiles and implementation profiles, as required for homeland
defense applications, for example, as well as for financial services,
health care, civil aviation, and the use of biometrics for preventing
identity theft.
The overall success of biometric systems depends on how well people who
use biometric systems accept them and how easy they are to use. If
enrollment and matching procedures are too cumbersome, data- capture
7 From 1997 to 2001, INCITS operated under the name Accredited Standards
Committee NCITS, National Committee for Information Technology Standards.
From 1961 to 1996, NCITS operated under the name Accredited Standards
Committee X3, Information Technology. User Acceptance
Chapter 3: Biometric Technologies for Personal Identification
Page 68 GAO- 03- 174 Biometrics for Border Security
errors can lead to high error rates, including FMRs and FNMRs. Moreover,
if people perceive a technology as being too intrusive, their lack of
cooperation or even resistance can affect a system*s performance. Privacy
concerns may be a barrier to the widespread adoption of biometric
technologies.
Some people find biometric technologies difficult, if not impossible, to
use. Still others resist biometrics in general as intrusive, inherently
offensive, or just uncomfortable to use. They consider it to be physically
intrusive to have to pause and position themselves in relation to a
capture device while presenting their biometric. Or they even consider
being required to verify their identity through a hardware device rather
than a human interaction to be too impersonal. Fingerprint systems, in
particular, face even stronger opposition because of their association
with criminal applications.
Some biometric devices also carry concerns about hygiene. For example,
some people object to hand geometry scanners because they do not like to
put their palms on the same surfaces where many other people have placed
theirs. Other people fear that devices that scan particularly sensitive
areas of the body, such as the eyes, will damage them. Generally, the less
intrusive people perceive a biometric to be, the more readily they accept
it.
Much public concern about biometrics arises from fears that the technology
can be misused to invade or violate personal privacy. Among these fears
are that biometric information will be
gathered without permission or knowledge or without explicitly defined
purposes,
used for a variety of purposes other than those for which it was
originally acquired (sometimes called *function creep*),
shared without explicit permission, or
used to track people across multiple databases to amalgamate information
for the purpose of surveillance or social control.
Chapter 3: Biometric Technologies for Personal Identification
Page 69 GAO- 03- 174 Biometrics for Border Security
No biometric technology is best for every situation, but it is possible to
determine which technologies are more accurate and easier to deploy for
border control applications. Last year, the International Civil Aviation
Organization (ICAO) assessed fingerprint, facial, and iris recognition as
the top three biometrics meeting the requirements for biometric
identification in machine- readable travel documents. Table 8 summarizes
the performance characteristics of the four technologies that are most
viable for border control. The performance factors such as error rates,
template sizes, and transaction times can vary greatly, depending on
whether the biometric technology is being used for 1: 1 verification or 1:
N identification.
Table 8: Four Viable Biometric Technologies Compared Characteristic Facial
Fingerprint Iris Hand
False nonmatch rate (FNMR) 3.3* 70% 0. 2* 36% 1. 9* 6% 0* 5% False match
rate (FMR) 0. 3* 5% 0* 8% Less than 1% 0* 2.1% User acceptance issues
Potential for privacy misuse Associated with law
enforcement; hygiene concerns
User resistance; usage difficulty Hygiene
concerns Enrollment time About 3 minutes About 3 minutes 30 seconds About
2 minutes 15
seconds About 1 minute Transaction time 10 seconds 9* 19 seconds 12
seconds 6* 10 seconds Template size 84* 1,300 bytes 250* 1, 000 bytes 512
bytes 9 bytes Number of major vendors 2 More than 25 1 1 Cost of device
Moderate Low High Moderate Factors affecting performance Lighting,
orientation of face, or
sunglasses Dirty, dry, or worn fingertips Poor eyesight, glare, or
reflections Hand injuries,
arthritis, or swelling Demonstrated vulnerability Notebook computer with
digital
photo or false photographs Artificial fingers or reactivated latent prints
High- resolution
picture of iris None Variability with age Affected by aging Stable Stable
Stable Commercially available 1990s 1970s 1997 1970s
Source: GAO analysis.
Recognizing that technology performance is least supported by substantive
real- life test data, ICAO has asked its member states to perform scenario
and operational evaluations with fingerprint, facial, and iris recognition
technologies. It plans to evaluate the results of the testing and to
select one or two biometric technologies for standardization in machine-
readable travel documents.
Retina, speaker, and signature recognition have certain drawbacks that
make them impractical for border control. Retina recognition is
Technologies Viable
for U. S. Border Control
Chapter 3: Biometric Technologies for Personal Identification
Page 70 GAO- 03- 174 Biometrics for Border Security
considered too intrusive because the systems require users to position
their eyes very close to devices, which some users find very
discomforting. Also, because using these systems requires prolonged effort
and concentration, a high percentage of people are unable to enroll.
Speaker recognition was piloted for border control use but has been found
unreliable. In fact, this technology has several disadvantages. Speech
quality is affected by a person*s health, such as a cold or sore throat,
stress, and emotions. In addition, speaker recognition systems do not
perform well in noisy environments because surrounding noise interferes
with their ability to extract the distinctive characteristics of an
individual*s speech. Moreover, because speaker recognition technologies
have large templates, they require longer processing times and use more
storage. Finally, the voice does not appear to be sufficiently distinctive
to permit identifying one individual within a large database of
identities. Signature recognition has a high FNMR because most people do
not sign their names consistently. Since the resulting nonmatches would
require many secondary inspections, signature recognition is probably not
practical for border control. Moreover, travelers from some countries may
not be accustomed to signing their names, to writing their names in roman
letters, or to writing at all.
The two leading vendors of facial recognition technology have their own
methods for analyzing a facial image and converting it to a digital
template. Enrolling in a facial recognition system seems relatively easy.
Results from Britain*s NPL product testing produced a 0 percent FTER. But
the performance of facial recognition technology appears to depend on the
operational setting and specific application. Pilots of facial recognition
surveillance at airports have resulted in FMRs between 0.3 percent and 5
percent and FNMRs between 5 percent and 45 percent. In a State Department
Bureau of Consular Affairs test involving data sets of 10,000 to 100,000
images, fewer than 30 percent of intentionally seeded duplicate images
were correctly matched* an FNMR of around 70 percent. Although facial
recognition performs much worse than fingerprint and iris recognition, it
remains attractive because facial images are used in a wide variety of
identification documents.
The performance of facial recognition technology is affected greatly by
environmental factors, especially lighting conditions. Variations in
camera performance and facial position, expression, and features
(hairstyle, eyeglasses, beards) further affect performance. Accurate image
alignment is necessary for the leading facial recognition algorithms,
which rely on identifying eye positions. One algorithm is rendered
ineffective when a Facial Recognition
Performance
Chapter 3: Biometric Technologies for Personal Identification
Page 71 GAO- 03- 174 Biometrics for Border Security
person tilts the head from a direct frontal pose to more than about 25
degrees horizontally or more than about 15 degrees vertically.
Performance is also degraded significantly as the stored facial
recognition template ages. When a match was attempted a year after initial
enrollment, some facial recognition technologies correctly verified as
little as 41 percent of the faces; this translates to an FNMR of 59
percent.
In tests conducted by the Federal Aviation Administration (FAA) from
November 2001 through January 2002, the average enrollment time was 3
minutes and 2 seconds. When the device was in use, the time increased by
approximately 9. 5 seconds to pass through a door.
Facial recognition systems can be quite costly. A facial recognition
server controlling access at a facility with up to 30,000 persons would
cost about $15,000. Depending on the number of entrances installed with
facial recognition devices, the cost of software licenses would range from
about $650 to $4,500. As the size of the database and the number of
attempted matches increased, so would a system*s cost. In addition to the
server and software licenses, a live- scan facial recognition surveillance
system includes closed- circuit television (CCTV) surveillance. A fully
integrated CCTV system for physical access surveillance can cost from
$10,000 to $200,000, depending on the size of the entrance and the degree
of monitoring required. For additional CCTV equipment, cameras can cost
between $125 and $500. Cameras with advanced features can cost up to
$2,300.
Although users typically consider facial recognition technology less
intrusive than other biometric technologies, some are concerned that it
can track them without their consent. Successful attempts to spoof
livescan facial recognition systems would not work in a border inspection
where a border inspector is monitoring the equipment. (See appendix IV for
more details on facial recognition technology.)
The majority of the leading vendors of fingerprint recognition technology
sell scanners based on optical or silicon technology. The companies*
techniques for converting a fingerprint image to a digital template are
proprietary. The basic performance of fingerprint recognition technology
depends on the type of application and the type of scanner capturing the
fingerprint image. For about 2 to 5 percent of people, fingerprints cannot
be captured because they are dirty or have become dry or worn from age,
extensive manual labor, or exposure to corrosive chemicals. Fingerprint
Recognition
Performance
Chapter 3: Biometric Technologies for Personal Identification
Page 72 GAO- 03- 174 Biometrics for Border Security
The time to enroll a person in a fingerprint recognition system depends on
the number of fingerprints used and the details of the enrollment process.
For example, in FAA testing, enrollment averaged 3 minutes and 30 seconds.
In contrast, in the first 7 months of the CANPASS* Airport pilot at
Vancouver International Airport, roughly 1,000 travelers registered in an
average of 15 minutes.
The time required to match a fingerprint and verify an individual*s
identity can vary from sensor to sensor and from one application to
another. For example, in FAA testing, users took an average of about 10
seconds to pass through the door, compared to an average of about 2
seconds before the device was installed. NPL found that an optical
fingerprint system had a mean transaction time of 9 seconds, while a
silicon sensor system had a time of 19 seconds.
A fingerprint recognition device can typically be set for different
security levels, with higher FMRs at lower levels of security. For
example, the FBI*s Integrated Automated Fingerprint Identification System
(IAFIS) has a 1.5 x 10 -12 FMR with an FNMR between 1.5 and 2 percent. In
contrast, FAA testing from September 2000 to February 2001 produced FNMRs
that ranged from about 6 percent to about 17 percent for closely
controlled test subjects. For actual airport employees accessing the door
in a lesscontrolled environment, the FNMR ranged from about 18 percent to
about 36 percent. The FMR ranged from 0 percent at the highest security
level to about 8 percent at the lowest security level.
The cost of each fingerprint reader designed for physical access control
ranges from about $1,000 to about $3,000. Software licenses are listed for
about $4 per enrolled user. For smaller fingerprint scanners, maintenance
is between 15 percent and 18 percent of cost. A larger live- scan 10-
print fingerprint reader costs about $25,000. Maintenance of the larger
machines is approximately 14 percent of the cost of the reader.
Because law enforcement agencies have used fingerprints to identify
criminals, the technology*s similarity to forensic fingerprinting causes
some people discomfort. Privacy advocates fear that fingerprint
recognition systems may collect data for one purpose but then use the data
to track people*s private activities or for other purposes. Also, people
may have hygiene issues with having to touch the plate of the scanner that
many other people have touched.
The fingerprint recognition technologies have been shown to be susceptible
to deception, but this can be prevented if fingerprints are
Chapter 3: Biometric Technologies for Personal Identification
Page 73 GAO- 03- 174 Biometrics for Border Security
scanned in a monitored environment. (See appendix II for more details on
fingerprint recognition technology.)
The sole provider of iris recognition technology developed the first
commercially viable system in 1997. Enrolling in an iris recognition
system requires a person to gaze steadily at a camera for a short time.
Some people find this difficult to do and therefore fail to enroll. The
FTER in an NPL test was 0.5 percent. While iris technology does not
require touching any device, some people resist the scanning of their
eyes.
However, iris recognition technology has good performance characteristics.
Testing at the U. S. Army Research Laboratory resulted in FMRs of less
than 1 percent and an FNMR of 6 percent. In 1996, Sandia National
Laboratories, testing a prototype iris recognition system, found that the
FNMR was 10.2 percent and the average enrollment time was 2 minutes and 15
seconds. In a more recent test by NPL, the iris recognition system showed
an FMR of 0 percent, FNMR of 0.2 percent, and a mean transaction time of
12 seconds.
Colored or bifocal contact lenses can affect system performance, as can
exceptionally strong glasses. Poor eyesight may also hinder some people
from lining their eyes up with the camera. Glare and reflection can also
cause interferences. People with glaucoma or cataracts may not be reliably
identified by iris recognition systems.
Iris recognition systems cost approximately $2,000 for physical access
units. The overall cost of a comprehensive iris recognition system would
be much higher.
Certain iris recognition devices have been spoofed by holding up to the
camera a high- resolution picture of an iris with a tiny hole cut out to
allow the pupil of a live eye to shine through. Such deceptions could be
prevented at a border inspection station monitored by inspectors. (See
appendix V for more details on iris recognition technology.)
Hand geometry, in use for almost 30 years, is a relatively mature
biometric technology with only one primary vendor. The shape and size of
our hands are reasonably diverse but not highly distinctive. Thus, hand
geometry is not suitable for identifying one individual among many.
Because border control applications require checking for duplicate
enrollment before travel documents are issued, hand geometry is not viable
for that aspect of Iris Recognition
Performance Hand Geometry Performance
Chapter 3: Biometric Technologies for Personal Identification
Page 74 GAO- 03- 174 Biometrics for Border Security
border control. However, hand geometry can be used to verify identity
after performing the enrollment checks with a more distinctive biometric
technology.
Typically, everyone with a hand can enroll in a system* FTER is 0 percent.
In FAA testing from March through July 2001, time for enrolling with a
hand geometry device averaged 57 seconds. The FNMR for airport employees
using the system ranged from approximately 5 percent at a high security-
level setting to less than 1 percent at a low security- level setting. The
FMR ranged from 0 percent at the high security- level setting to about 2
percent at the low security- level setting. The FAA test also found that
using the hand geometry device increased the time to open a door by 6
seconds. However, an NPL test found a mean transaction time of 10 seconds
for a hand geometry system. The performance of hand geometry technology is
affected by jewelry, arthritis, water retention, and swelling from
pregnancy or hand injury.
Hand geometry devices generally cost between $2,000 and $4,000. Staff
training is minimal, with no personnel costs, since most hand geometry
devices are unattended. It is considered easy to use, although a minimal
amount of training may be required for individuals to learn to align their
hands in the device. Hand geometry is generally perceived as not
intrusive, not threatening, and not invasive, and it bears very little of
the stigma of other biometric technologies. (See appendix III for more
details on hand geometry technology.)
Applying biometric technologies to customs and immigration in the United
States and other nations is growing rapidly. Fingerprint, facial, and iris
recognition and hand geometry systems are being planned or have been
implemented to different degrees, ranging from piloted tests to
operational usage. We summarize some of these projects and their
applications, particularly to trusted air travel, land border crossing,
obtaining and verifying travel documents, and surveillance.
Trusted air travel programs permit frequent travelers to circumvent
customs procedures and immigration lines. To participate, users undergo a
background screening and registration. Once enrolled, they can present
their biometric at an airport kiosk for comparison against a template
stored either on a storage card in their possession or in a central
database. Biometric Technology
Applied to Border Control Today
Trusted Air Travel
Chapter 3: Biometric Technologies for Personal Identification
Page 75 GAO- 03- 174 Biometrics for Border Security
INSPASS, a pilot program in place since 1993, has more than 35,000
frequent fliers enrolled at nine airports, with more than 250,000
transactions every year. It is open to citizens of the United States,
Canada, Bermuda, and visa waiver program countries who travel to the
United States on business three or more times a year.
A hand recognition system similar to INSPASS at Ben Gurion Airport in Tel
Aviv, Israel, since 1998 verifies international travelers and all Israeli
citizens. By April 2002, more than 100,000 travelers had enrolled in the
program, and the system was processing about 50,000 passengers each month.
The Expedited Passenger Processing System (EPPS), based on iris
recognition technology, is being launched at eight major international
airports in Canada. Positive verification against the template at an
airport kiosk entitles travelers to circumvent customs and immigration
lines. The first kiosks are expected to be installed in Vancouver and
Toronto airports in 2003.
In July 2001, frequent travelers on British Airways and Virgin Atlantic
Airways transatlantic flights began clearing immigration through iris
recognition verification at London*s Heathrow Airport. Once registered and
enrolled, landing passengers can proceed directly to special lanes to
verify their identity against an iris template stored in a central
database. If successful, they are issued a ticket that admits them
directly to the United Kingdom.
A program to expedite immigration processing for frequent travelers at
Amsterdam*s Schiphol Airport, the Netherlands, is based on a combination
of iris recognition and smart card technology. About 2,000 smart cards
have been issued to nationals from 18 different European countries.
In a joint INS and State Department effort to comply with the Illegal
Immigration Reform and Immigrant Responsibility Act of 1996, every border
crossing card issued after April 1, 1998, contains a biometric identifier
and is machine- readable. The cards, also called laser visas, allow
Mexican citizens to enter the United States without being issued further
documentation for the purpose of business or pleasure and stay for 72
hours or less, going no farther than 25 miles from the border. If a
Mexican citizen plans to stay for longer than 72 hours or to go more than
25 miles from the border, additional documentation is required. Consular
staff in Mexico photograph applicants and take prints of the two index
fingers and Land Border Crossing
Chapter 3: Biometric Technologies for Personal Identification
Page 76 GAO- 03- 174 Biometrics for Border Security
then electronically forward applicants* data to INS. Both State and INS
conduct checks on each applicant, and the fingerprints are compared with
prints of previously enrolled individuals to ensure that the applicant is
not applying for multiple cards under different names. The cards store a
holder*s identifying information along with a digital image of his or her
picture and the minutiae of the two index fingerprints. Figures 13 and 14
show the front and back of the laser visa. As of May 2002, State had
issued more than 5 million cards. However, INS has not yet deployed
fingerprint readers or card readers, so inspectors examine cards at the
points of entry as they would a travel document.
Figure 13: The Front of a Laser Visa
Source: INS.
Chapter 3: Biometric Technologies for Personal Identification
Page 77 GAO- 03- 174 Biometrics for Border Security
Figure 14: The Back of a Laser Visa
Source: LaserCard Systems Corporation, Mountain View, California.
The government of Israel is implementing a biometrics system that uses
hand and facial scans to facilitate passage through border checkpoints
between the Gaza Strip and other areas of Israel. The system will verify
the identity of 60,000 Palestinian workers who cross the border at 42
automated checkpoints daily. The workers* biometrics will be compared with
templates stored on a central server and backed up on smart cards that the
workers can present.
An iris recognition system in Singapore processes motorbike passengers
crossing the border from Malaysia each day to work. Approximately 50,000
travelers cross this border each day.
Hong Kong plans to introduce a fingerprint scanning system in 2003 at the
Shenzhen border in China to accelerate immigration for the 250,000 people
who cross the border every day. Travelers will be able to swipe a smart
card bearing personal data along with a photograph and the template of a
thumbprint through an optical reader while presenting the thumb to a
scanner.
The Department of State has been running pilots of facial recognition
technology at 23 overseas consular posts for several years. As a visa
applicant*s information is entered into the local system at the posts and
Obtaining and Verifying
Travel Documents
Chapter 3: Biometric Technologies for Personal Identification
Page 78 GAO- 03- 174 Biometrics for Border Security
replicated in State*s CCD, the applicant*s photograph is compared with the
photographs of previous applicants stored in CCD to prevent fraudulent
attempts to obtain visas. Some photographs are also being compared to a
watch list.
Australia*s Sydney Airport is assessing facial recognition technologies in
one- to- one comparisons of individuals* facial features with their
passport pictures to identify people traveling with false passports.
Saudi Arabia installed iris scanning and fingerprinting devices in the
King Abdul Aziz Airport in the Red Sea port city of Jeddah during this
year*s annual Hajj pilgrimage to Mecca to verify that individuals entered
and exited the country under the same travel documents.
Sydney Airport in Australia is using facial recognition technology to
identify wanted faces within the airport*s crowds. Iceland*s main
international airport at Keflavik scans passengers with facial recognition
technology as they pass through boarding gates, comparing their facial
characteristics with a watch list of suspected terrorists and criminals.
Surveillance
Chapter 4: Scenarios for Border Control with Biometrics
Page 79 GAO- 03- 174 Biometrics for Border Security
In the previous chapter, we described how biometric technologies work,
their performance, and some of their applications. In this chapter, we
outline how fingerprint, facial, and iris recognition technologies could
help improve the procedures now used to secure U. S. borders. We identify
four possible scenarios:
Making a watch list check before issuing travel documents.
Making a watch list check before travelers enter the United States.
Issuing U. S. visas with one or more of these biometrics.
Issuing U. S. passports with one or more of these biometrics. These
scenarios do not represent all the ways to use biometrics for border
control, but they do reflect some elements of pilots that have implemented
biometric technologies for border control, as well as options discussed in
legislation and by agencies responsible for border security. While hand
geometry cannot be used for conducting a watch list check, it can be used
in conjunction with one of the other technologies to verify identities
using visas or passports.
The first two scenarios could help identify individuals who are ineligible
to receive a U. S. visa or passport or who cannot be admitted to the
United States. Both of these scenarios use an identification match to
compare the traveler*s biometric against a database of stored biometrics.
The two other scenarios could help link an individual*s identity to U. S.
travel documents and could help reduce document counterfeiting and
impostors* fraudulent use of legitimate documents. The four scenarios are
not mutually exclusive; they could be implemented individually or in
combination. In the next chapter, we analyze costs, benefits, and
implications associated with implementing the scenarios.
Making a watch list check before issuing travel documents could identify
individuals ineligible to receive a U. S. visa or passport when their
biometric was compared during the application process against a database
of the biometrics of individuals on a watch list. This scenario would have
the least effect on current operations and would require the least
development of new systems. As depicted in figure 15 for visas and figure
16 for passports, the watch list check would essentially be an additional
computer check conducted much as the name check that is conducted through
CLASS today. Chapter 4: Scenarios for Border Control with
Biometrics Watch List Check before Issuing Travel Documents
Chapter 4: Scenarios for Border Control with Biometrics
Page 80 GAO- 03- 174 Biometrics for Border Security
Figure 15: Issuing U. S. Visas by a Watch List Check Process
Source: GAO analysis. Submits Application
1 photo
Passport 210 American embassies and consulates
Mail or drop box
Name check
Multiple visas
Biometric watch list check
Possible security advisory opinion required
State HQ Visa issued? Option
Decides Visa
or Visa Consular officer
Applicationpackage
Computer checks
Interview by consular
officer
Chapter 4: Scenarios for Border Control with Biometrics
Page 81 GAO- 03- 174 Biometrics for Border Security
Figure 16: Issuing U. S. Passports by a Watch List Check Process
Source: GAO analysis.
Policies for the contents of the watch list would have to be developed,
including criteria for names to place on the watch list* whether those of
terrorists, criminals, violators of immigration law, or others. The
biometric technology would most likely be based on facial recognition from
Passport Passport To renew
2 photos
Application
Old passport
Name check
Multiple passports
Multiple applications
Biometric watch list check
Passport examiner 16 passport centers
Decides Mail
or Send
results To apply
2 photos
Application
Proof of citizenship
Proof of identity More than 4,500 passport offices
Applicationpackage -OR Passport issued?
Computer checks
Passport acceptance agent
Central application processing
center Electronically keyed into U. S.
State Department*s
computer system Application
package
Chapter 4: Scenarios for Border Control with Biometrics
Page 82 GAO- 03- 174 Biometrics for Border Security
photographs that applicants for documents submit. Often, a photograph is
the only biometric available for certain people who are not admissible to
the United States. Criteria for the quality of the stored biometric for
those on the watch list would probably have to be developed in order to
enhance the performance of the matching process.
Implementing this scenario would probably require two additional computer
system units to house the watch list and to match applicants* photographs
and the photographs on the watch list. Figure 17 depicts one possible
construct for this scenario*s architecture. Existing systems, such as CCD
and Passport Files Miniaturization (PFM), could require significant
changes and corresponding time and resources to accommodate this scenario.
Figure 17: System Architecture for a Biometric Watch List Check before
Issuing Travel Documents
Source: GAO analysis.
Depending on the watch list criteria, it might be possible to use
fingerprint or iris recognition to perform the match. Using fingerprints
or the iris as the watch list biometric would complicate data collection.
Instead of just submitting a photograph, applicants would have to submit
fingerprint or iris biometrics. This information would then have to be
stored centrally and read by readers installed at embassies, consulates,
and passport acceptance offices.
Passport Files Miniaturization
Facial recognition
engine Watch list photograph
database Consular
Consolidated Database
Nonimmigrant visa system
Consular Lookout and
Support System Passport
Records Imaging
System Management
Passport centers
Headquarters Embassies and consulates
Chapter 4: Scenarios for Border Control with Biometrics
Page 83 GAO- 03- 174 Biometrics for Border Security
Individuals who are not eligible to enter the United States could be
identified before they could enter if, during inspection, their biometrics
are checked against a database of the biometrics of people on a watch
list. As depicted in figure 18, this watch list check* which would be
similar to the IBIS check at ports of entry* would be an additional
computer check conducted during inspection.
Figure 18: Entering the United States by a Watch List Check Process
Source: GAO analysis.
Watch List Check before Entering the United States
If problem - secondary inspection
Thorough interviews
Comprehensive computer checks
Primary inspection J. DOE ? AOK- 11
If no problem Does admit If problem
Does not admit Document
inspection Name check Vehicle license plate check (land only)
Interview Biometric watch list check
Chapter 4: Scenarios for Border Control with Biometrics
Page 84 GAO- 03- 174 Biometrics for Border Security
As with the watch list scenario for issuing travel documents, policies
would have to be developed for a watch list for entering the country,
including the list*s contents and the quality of the stored biometric.
Facial recognition based on images collected as travelers presented
themselves before INS inspectors would be the likely biometric technology.
Often, a photograph is the only biometric available for certain people not
admissible to the United States.
As with the scenario we described above, a database to store the watch
list would have to be developed. The primary difference in cost between
these two scenarios would be the cost of biometric readers for the ports
of entry and the corresponding infrastructure and personnel to use the
readers. The readers would require access to the database of the
biometrics of the individuals on the watch list. Figure 19 depicts one
possible construct for this scenario*s architecture. Existing systems,
such as IBIS and the Treasury Enforcement Communications System (TECS),
could require significant changes and corresponding time and resources to
accommodate this scenario.
Figure 19: System Architecture for a Biometric Watch List Check before
Entering the Country
Source: GAO analysis. Facial
recognition engine
Treasury Enforcement Communications
System Interagency
Border Inspection
System Facial
recognition camera system Watch list
photograph database
Headquarters Ports of entry
Chapter 4: Scenarios for Border Control with Biometrics
Page 85 GAO- 03- 174 Biometrics for Border Security
In a scenario in which U. S. visas contained biometrics, two of the border
control processes would be affected. First, applicants for U. S. visas
would submit a biometric with their applications at American embassies and
consulates. During the application process, the applicant*s biometric data
would be stored and an identification match would be conducted to compare
the biometric information stored from other issued visas, as well as
rejected visa applications, to check for duplicate and fraudulent
applications. Second, at the ports of entry, the traveler*s biometric
would be verified as a part of the inspection process. The verification
match compares the biometric data collected during the visa application
process with the data collected during the inspection process.
Figure 20 shows how collecting the biometric would change current visa
issuing procedures and the additional computer check necessary to
determine whether the new biometric had been previously enrolled. Figure
21 shows how port of entry inspection would change* essentially by adding
a computer check to confirm travelers* identities. U. S. Visas with
Biometrics
Chapter 4: Scenarios for Border Control with Biometrics
Page 86 GAO- 03- 174 Biometrics for Border Security
Figure 20: Issuing U. S. Visas with Biometrics
Source: GAO analysis. Submits Application
1 photo
Passport
Biometric sample 210 American embassies and consulates
Mail or drop box
Name check
Multiple visas
Biometric check
Possible security advisory opinion required
State HQ Visa issued? Option
Decides Visa
Visa Consular officer
Applicationpackage
Computer checks
Interview by consular
officer
Chapter 4: Scenarios for Border Control with Biometrics
Page 87 GAO- 03- 174 Biometrics for Border Security
Figure 21: Port of Entry Visa Inspection with Biometrics
Source: GAO analysis.
This scenario would require buying biometric readers for the embassies,
consulates, and ports of entry. A database would be required for storing
biometric information. This database could be integrated with CCD, which
stores visa application and issuance information. To properly link a
biometric with an individual, live capture of the biometric would be
required, eliminating some, if not all, of the benefit of mail- in and
drop- box visa applications. Figure 22 shows one possible construct for
this
If problem - secondary inspection
Thorough interviews
Comprehensive computer checks
Primary inspection J. DOE ? AOK- 11
If no problem Does admit If problem
Does not admit Document
inspection Name check Vehicle license plate check (land only)
Interview Biometric check
Chapter 4: Scenarios for Border Control with Biometrics
Page 88 GAO- 03- 174 Biometrics for Border Security
scenario*s architecture. Fingerprint, facial, or iris recognition could be
used for this scenario. Hand geometry can be used only in combination with
another technology because it is not effective in identification matches.
Existing systems, such as IBIS, TECS, and CCD, could require significant
changes and corresponding time and resources.
Figure 22: System Architecture for Issuing Visas with Biometrics
Source: GAO analysis.
Two border control processes would be affected also in a scenario that
issued U. S. passports containing biometrics. First, passport applicants
would submit a biometric with their applications. The applicant*s
biometric data would be stored and an identification match would be
conducted to compare the biometric information stored from other issued
passports, as well as rejected passport applications, to check for
duplicate and fraudulent applications. Second, the traveler*s biometric
would be verified during inspection at ports of entry. The verification
match compares the biometric data from the passport application with the
data collected during inspection. Figure 23 shows the biometric collection
and the computer check to determine whether travelers* biometrics had been
previously enrolled. Figure 24 shows the port of entry inspection,
essentially adding a computer check to confirm travelers* identities. U.
S. Passports with
Biometrics
Treasury Enforcement Communications
System Biometric indentification
engine Interagency
Border Inspection
System Biometric
scanner Consular
Lookout and Support System
Biometric scanner Nonimmigrant
visa system Consular
Consolidated Database
Biometric storage
Headquarters Ports of entry
Embassies and consulates
Chapter 4: Scenarios for Border Control with Biometrics
Page 89 GAO- 03- 174 Biometrics for Border Security
Figure 23: Issuing U. S. Passports with Biometrics
Source: GAO analysis. Central
application processing
center Electronically keyed into U. S.
State Department*s
computer system
Passport Passport To renew
2 photos
Application
Old passport
Name check
Multiple passports
Multiple applications
Biometric check
Passport examiner 16 passport centers
Decides Mail
or Send
results To apply
2 photos
Application
Proof of citizenship
Proof of identity More than 4,500 passport offices
Application package
Applicationpackage
Biometric sample -OR Passport issued?
Computer checks
Passport acceptance agent
Chapter 4: Scenarios for Border Control with Biometrics
Page 90 GAO- 03- 174 Biometrics for Border Security
Figure 24: Port of Entry Passport Inspection with Biometrics
Source: GAO analysis.
This scenario would require purchasing biometric readers for passport
acceptance offices and ports of entry. It would require the database for
storing biometric information. The database could be integrated with the
State Department*s Passport Records Imaging System Management (PRISM) and
PFM, which store passport application and issuance information. To
properly link biometrics with individuals, live capture of biometrics
would be required, and this might eliminate some of the
If problem - secondary inspection
Thorough interviews
Comprehensive computer checks
Primary inspection J. DOE ? AOK- 11
If no problem Does admit If problem
Does not admit Document
inspection Name check Vehicle license plate check (land only)
Interview Biometric check
Chapter 4: Scenarios for Border Control with Biometrics
Page 91 GAO- 03- 174 Biometrics for Border Security
benefits of mail- in renewal applications. Figure 25 shows one possible
construct for this scenario*s architecture. Fingerprint, facial, or iris
recognition could be used for this scenario. Hand geometry can be used
only in combination with one of the other technologies because it is not
effective in performing identification matches. Existing systems, such as
IBIS, TECS, and PFM, could require significant changes and corresponding
time and resources to accommodate this scenario.
Figure 25: System Architecture for Issuing Passports with Biometrics
Source: GAO analysis.
Two or more of these scenarios could be implemented in combination.
Implementing scenarios in combination would not necessarily mean that
costs would be additive. For example, the same biometric readers could be
used to read biometrics from visa holders and passport holders at U. S.
ports of entry. Similarly, the same watch list database could be used for
checking before issuing travel documents and for checking before allowing
entry into the United States.
It would also be possible to implement multiple biometric technologies.
For example, it might be desirable for performance reasons to have both
facial and fingerprint biometrics captured on visas so that either or both
could be verified when people seek entry to the United States. It might be
Implementing
Multiple Scenarios
Biometric scanner
Passport Files Miniaturization
Treasury Enforcement Communications
System Interagency
Border Inspection
System Biometric
storage Biometric
identification engine
Biometric scanner Passport
Records Imaging
System Management
Passport acceptance
offices Headquarters
Ports of entry
Chapter 4: Scenarios for Border Control with Biometrics
Page 92 GAO- 03- 174 Biometrics for Border Security
possible to integrate the match algorithms so that they take in results
from both biometric readers and use them in combination to determine
matches. The incremental costs associated with the additional biometric
readers would have to be considered, as well as the costs of any
additional labor and space required in order to capture the biometrics and
any additional server capacity to store the additional biometrics. We
discuss costs in the next chapter.
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 93 GAO- 03- 174 Biometrics for Border Security
While biometric technology is currently available and used in a variety of
applications, questions remain regarding the technical and operational
effectiveness of biometric technologies in applications as large as border
control. In addition, before implementing any biometric border security
system, a number of other issues would have to be considered, including
The system*s effect on existing border control procedures and people.
Technology is only part of an overall security solution and only as
effective as the procedures within which it operates.
The costs and benefits of the system, including secondary costs
resulting from changes in processes or personnel to accommodate the
biometrics.
The system*s effect on privacy, convenience, and the economy. In this
chapter, we present our analysis of the costs and benefits of the four
scenarios as they could be applied to current border control procedures.
Ideally, a biometric should be universally present, unique to the
individual, and stable over time. The cost and ease of using a biometric
technology also weigh into its selection. Of the four biometrics we
examine in depth for border control, only a person*s face is universally
present, while other biometrics are not* people can lose or damage
fingers, hands, and eyes. Estimates are that 1 to 3 percent of the
population might be physically unable to use these biometrics.
Hand geometry and fingerprint, facial, and iris recognition have not been
formally proven unique. Therefore, a biometric*s uniqueness within a large
population can be established only by its historical use. Table 9 shows
the sizes of some of the larger biometric systems. Chapter 5: Applying
Biometrics to Border
Control: Challenges and Implications The Performance of Biometric
Technologies
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 94 GAO- 03- 174 Biometrics for Border Security
Table 9: The Enrollment Size of Seven Operational Biometric Systems
Biometric database Technology Enrollment
Mexican Federal Electoral Institute Facial recognition 60,000,000
Integrated Automated Fingerprint Identification system Fingerprint
40,000,000 INS Automated Biometric Fingerprint Identification System
Fingerprint 4, 500,000 Ben Gurion International Airport Hand geometry
100,000 INS Passenger Accelerated Service System Hand geometry 35,000 King
Abdul Aziz Airport, Saudi Arabia Iris recognition 30,000 Schiphol Airport,
Amsterdam Iris recognition 2,000
Source: GAO analysis.
As table 10 shows, the sizes of the biometric systems specified in the
four scenarios are large. The system required to issue visas with
biometrics far exceeds in size the largest biometric database created so
far. While fingerprint and facial recognition have been used in large
systems, the size of each scenario far exceeds the largest iris
recognition system of 30,000. As we have previously described, hand
geometry is not highly distinctive and therefore cannot be used for border
control where identification of one individual among many will be
required. To be used for verification, hand geometry will need to be used
in combination with one of the other technologies that can perform the
initial identification match.
Table 10: Estimated Number of Biometric Matching Transactions in Four
Border Control Scenarios Scenario System size Matching transactions
per year
1. Making a watch list check before issuing travel documents Depends on
criteria used to develop the watch
list: CLASS has about 10 million records for foreigners and U. S. citizens
17 to 31 million applications:
10 million visas;
7 million passports;
possibly 14 million visas from visa waiver countries 2. Making a watch
list check before travelers enter the United States Depends on criteria
used to develop the watch
list: CLASS has about 10 million records 500 million primary inspections
3. Issuing U. S. visas with biometrics 100 million to 240 million visa
records over 10
years 48 to 63 million 4. Issuing U. S. passports with biometrics 70
million passport records over 10 years Up to 175 million a
a About 175 million U. S. citizens were inspected at ports of entry in
fiscal year 2001. Because a passport is not required in returning from
countries such as Canada and Mexico, it is not clear how many of these
citizens had passports.
Source: GAO analysis.
In testing and operation, some fingerprint and iris recognition
technologies have proven fairly accurate. Fingerprint recognition has
achieved a low FMR but a variable FNMR. According to the FBI, the FMR for
IAFIS is about 1.5 x 10 -12 with an FNMR of between 1.5 and 2.0 percent.
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 95 GAO- 03- 174 Biometrics for Border Security
In testing NPL conducted, at an FMR of about 2 percent, the FNMR was about
4.3 percent. In pilots FAA sponsored, FNMR ranged from 6 percent to 36
percent and the FMR was between 0 percent and 8 percent.
Iris recognition has also shown it can achieve a low FMR but with a
variable FNMR. In NPL*s testing, the FMR was 0 percent with an FNMR of 1.9
percent. The U. S. Army Research Laboratory found an FMR below 1 percent
with an FNMR of 6 percent. Sandia National Laboratories* test showed 0
percent FMR and 10.2 percent FNMR.
Facial recognition has had more mixed results. In verification testing NPL
conducted, at an FMR of about 1 percent, the FNMR was about 3.3 percent.
In a pilot FAA sponsored, an FMR of 0.19 percent was achieved with an FNMR
between 3 percent and 26 percent. In preliminary testing NIST conducted
this year, facial recognition achieved an FMR of 1 percent and an FNMR of
25 percent. For identification testing, facial recognition fared worse. A
State Department pilot encountered an FMR of 15 percent. Tests conducted
at U. S. airports have found FMRs between 1 and 5 percent and FNMRs
between 5 and 15 percent. At one airport where the objective was to
achieve an FMR as close to 0 as possible, an FMR of 0.3 percent was
achieved but with an FNMR of 45 percent. The U. S. Army Research
Laboratory found an FMR of 49 percent.
The final primary factor to consider when evaluating biometrics is
stability over time, but little work has been done to establish this.
Fingerprints are believed to be persistent from birth throughout life. It
is believed that irises are stable from before birth until death. FRVT
2000 tested facial recognition with images collected a year before
identification or verification. The FMR for verification was 44 to 59
percent, while for identification it was 52 to 69 percent.
Fingerprint recognition appears to be the most mature of these biometric
technologies. Fingerprint recognition has been used the longest and has
been used with databases containing up to 40 million entries. Iris
recognition is young and has not been used with populations approaching
the size entailed in border control. While facial recognition has also
been used with large databases, its accuracy results in testing have
lagged behind those of iris and fingerprint recognition. IBG believes that
further research, costing between $50 million and $100 million, would be
required to determine whether iris or facial recognition could perform at
the same level as fingerprint recognition.
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 96 GAO- 03- 174 Biometrics for Border Security
The success of any border security technology depends on the border
control procedures as well as the people engaged in those procedures.
Technology is not a solution in isolation. Technology and people must work
together to execute the border control process* from issuing travel
documents to inspecting them at official ports of entry.
Introducing biometrics would affect people and processes differently,
depending on the specific scenario. Further, the performance of the
biometric technology can also affect the overall process. To check a watch
list before issuing travel documents, the following would need to be
considered:
Installation of readers at consulates and embassies for visa operations
and at passport acceptance offices for passport operations would require
hiring additional staff and, in some cases, leasing additional space.
While the watch list identification check is essentially just an
additional computer check, high FMRs could increase the work of consular
officers and passport examiners and could delay the disposition of
applications if significant time were required to reconcile false hits.
Consular staff, passport acceptance agents, and passport examiners would
have to be trained.
Mail- in and drop box applications could be expected to fall off
considerably, if not completely.
Similar concerns would need to be addressed to check a biometric watch
list before travelers enter the country.
Installing readers at ports of entry would require hiring additional
staff and, in some cases, leasing additional space.
Because the watch list identification check is essentially just an
additional computer check, similar to an IBIS check, hits would probably
result in secondary inspection of the traveler. High FMRs could increase
the work of inspectors and delay the passage of travelers if significant
time were required to reconcile false hits.
Inspectors would have to be trained to collect the biometric from
travelers and to resolve watch list hits in secondary inspection. An
outreach campaign would likely be necessary to educate travelers about the
new biometric program. How Introducing the
Technology Affects People and Procedures
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 97 GAO- 03- 174 Biometrics for Border Security
One key impact, the increased time required to conduct an inspection with
a biometric watch list, would result from three key factors. First, to
check all identities through IBIS using a biometric watch list would be a
more substantive security check that would lengthen primary inspection. As
we have previously described, not all travelers are now subjected to an
IBIS name check. Second, while some have suggested that biometrics could
speed inspection, FAA tests suggest biometrics would slow it down. FAA
tests with biometric technology in a physical access environment showed
that transit time increased by 6 to 9 seconds when biometrics were added
to a magnetic card entry system. Third, an FMR that is too high could lead
to excessive referrals of travelers to secondary inspection and could
increase workload to resolve the false matches. For example, using facial
recognition with a watch list of 10 million people and just a 1 percent
FMR would result in an average of 100,000 false matches per traveler.
Clearly, if the watch list will be large, the FMR will need to be
extremely low to maintain workload at a manageable level.
For both watch list scenarios, policies and procedures would have to be
developed for adding and maintaining records in the watch list database.
Key questions that have to be answered for a watch list database include
who is added to the watch list, how someone is removed from the watch
list, and how errors can be corrected. One of the biggest issues would be
the selection of a biometric to identify individuals on the watch list.
Today*s watch lists are primarily name- based and frequently list only the
individual*s name, approximate age, suspected nationality, or other
identifying data. The selection could be affected by who will be placed
into the watch list because biometric information for some people is not
available. Facial recognition could be the likely biometric technology for
a watch list because often only photographs are available for certain
people inadmissible to the United States. However, fingerprint recognition
or iris recognition could also be used if the United States could collect
records on those individuals.
To issue and verify visas with biometrics, changes would be required at
embassies and consulates to issue the visas and at ports of entry to
verify the identities of those traveling with visas. Specifically, the
following would need to be considered:
Installing readers at consulates and embassies for visa operations would
require hiring additional staff and, in some cases, leasing additional
space.
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 98 GAO- 03- 174 Biometrics for Border Security
While the biometric identification check for duplicate or rejected
applications is essentially just an additional computer check, high FMRs
could increase the work of consular officers and delay the disposition of
visa applications if significant time were required to reconcile false
hits.
Consular staff would have to be trained.
Mail- in and drop box applications could be expected to fall off
considerably, if not completely.
Similarly, to issue and verify passports with biometrics, passport
acceptance office operations could be dramatically modified. Because the
vast majority of these offices are not State Department offices and do not
have State Department personnel or equipment, policy decisions would have
to be made regarding the installation of computers and biometric equipment
at these offices. Specifically, the following would need to be considered:
Installing readers at passport acceptance offices would require hiring
additional staff and, in some cases, leasing additional space.
Because there is not a State Department presence at passport acceptance
offices, a mechanism would need to be developed to transmit the collected
biometrics on removable media or through a network connection to the
department.
While the biometric identification check for duplicate or rejected
applications is essentially just an additional computer check, high FMRs
could increase the work of passport examiners and could delay the
disposition of passport applications if significant time were required to
reconcile false hits.
Passport acceptance agents and passport examiners would have to be
trained.
Mail- in applications could be expected to fall off considerably, if not
completely.
As we previously described for the use of a biometric watch list at the
ports of entry, the use of biometrics with visas or passports would likely
lengthen the inspection time. Although the matching operation conducted
with visas or passports with biometrics would be a verification match
instead of an identification match, the inspection time could still go up
for
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 99 GAO- 03- 174 Biometrics for Border Security
the same reasons. Checking that the bearer of a travel document is the
proper bearer of the document is a more stringent check than is conducted
today. Further, the performance of the biometric technology affects the
number of secondary inspections conducted if travelers are not properly
matched to their biometric. Other issues that would need to be considered
include
Installing readers at ports of entry would require hiring additional
staff and, in some cases, leasing additional space.
Because the biometric verification check is essentially just an
additional computer check, similar to an IBIS check, hits would probably
result in secondary inspection of the traveler. An FMR that is too high
could lead to inadmissible people being allowed to enter the country. An
FNMR that is too high could lead to an increase in the number of travelers
referred to secondary inspection, adding to requirements for space and
personnel.
Inspectors would have to be trained to collect the biometric from
travelers and to resolve watch list hits in secondary inspection. An
outreach campaign would likely be necessary to educate travelers about the
new biometric program.
The biometrics for visas and passports could be stored and verified with
or without tokens. Biometric data could be stored on tokens travelers
carried, to be compared with data from biometric readers at ports of
entry. A token could be a traveler*s visa or passport with the biometric
data stored on it as a bar code, or it could be a separate memory storage
card, such as a smart card or laser card.
In an approach without tokens, a traveler*s biometric data would be stored
in a central database to be queried during matching. The data in the
central database could be indexed by the visa or passport number or simply
by the traveler*s name combined with other identifying information such as
date of birth, Social Security number, or driver*s license number.
Regardless of the comparison method for verification, the enrollment
process would be the same, whether at a consulate, embassy, or passport
acceptance office. It is critical that the biometric, once collected, be
securely linked to the visa or passport application and stored in a
central database for comparison to other records, ensuring that duplicate
identities are not being created. The operational concepts are
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 100 GAO- 03- 174 Biometrics for Border Security
Check against token containing biometric data. The traveler enters a
primary inspection area and presents to the inspector a token containing
his or her biometric data. The token is read and the biometric data are
decrypted and validated. The traveler*s stored biometric data and the
biometric data obtained from the biometric reader are compared. If the
data match, and if the inspector has no other reason to deny admission,
then the traveler is admitted to the United States.
Check against central database for biometric data. The traveler enters a
primary inspection area and presents to the inspector a travel document or
some other identifying information. Lacking a visa or passport, the
traveler must provide information detailed such that a single record can
be pulled from the central database. The remaining steps are the same as
in checking a biometric token.
Process flow issues must be considered. A central database of biometrics
would be required to prevent people from getting multiple passports or
visas under different identities and for verifying the identity of a
traveler whose token has been lost or stolen or becomes unusable. In this
case, it is important that the traveler be able to provide enough
information so that the inspector can check for and find the appropriate
records. It is also possible that an identification match, instead of a
verification match, could be run on an individual.
If a token is used, how it is produced must be considered. If it is to be
a modification of the current passport or visa* for example, if the
biometric is a two- dimensional bar code stored on the travel document*
redesigning the passport or visa foil would be required. If the token is
to be a separate card, such as a smart card or a laser card, the capital
investment in a production facility would have to be considered.
Using tokens for the biometric storage could affect the inspection
process. No studies have yet determined whether tokens expedite
inspection. Studies should be conducted to determine what effect local
data comparisons would have, compared with central database lookups.
For the three scenarios with biometric scanners at ports of entry, the
physical configuration at the ports of entry could pose a challenge for
collecting travelers* biometrics and performing matches. Where there are
terminals, such as at airports, some seaports, and pedestrian ports of
entry, it would be relatively simple to install biometric readers and to
read travelers* biometrics. An inspector checks travelers* identities and
names
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 101 GAO- 03- 174 Biometrics for Border Security
at booths equipped with IBIS. At most sea ports of entry, where IBIS is
not used, inspectors board the vessels to conduct inspections of aliens,
while U. S. citizens are inspected as they disembark. Biometric readers
could not be installed in such circumstances, making collecting biometrics
from travelers challenging. Similarly, at land ports of entry, a way to
collect biometrics expeditiously from all occupants of a vehicle would
have to be developed.
For all four scenarios, exception processing would have to be carefully
planned. When an applicant fails to enroll in a biometric system or when a
system fails to correctly identify a person, that person must be treated
as an exception. Exception processing that is not as good as biometric-
based primary processing could be exploited as a security hole. Exceptions
include passport and visa applicants whose biometrics cannot be properly
enrolled in the system because they may not have the physiological
characteristic that the system recognizes. One solution might be to use
two or more biometric technologies in the same system, reducing the number
of people who could fail to be enrolled.
The failure of biometric scanners, failure to access the central biometric
database, failure to access the watch list, and communications failure are
other exceptions. Because it is unlikely that inspections would cease,
appropriate contingency plans would have to be developed to ensure
continuity of operations without sacrificing security. Further, an
appropriate transition strategy will be required to handle simultaneously
biometric travel documents and the current travel documents that will
remain valid without biometrics for the next 10 years.
Just as operational processes must be considered, infrastructure processes
must also be examined, particularly with respect to information security.
Binding an identity to the biometric features of a person is only an entry
in a database. Lax information security can weaken or break that bond.
Laws enacted over the past 15 years require each federal agency to provide
security protections for information collected and maintained by or for
the agency commensurate with the risk and magnitude of harm that
Biometrics and
Information Security
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 102 GAO- 03- 174 Biometrics for Border Security
would result from unauthorized disclosure, disruption, modification, or
destruction of the information. 1
Despite these statutory requirements, we have previously reported that
poor information security is a widespread federal problem with potentially
devastating consequences. 2 Although agencies have taken steps to redesign
and strengthen their information system security programs, our analyses of
information security at major federal agencies have shown that federal
systems were not being adequately protected from computer- based threats,
even though these systems process, store, and transmit enormous amounts of
sensitive data and are indispensable to many federal agency operations.
These weaknesses continue, as indicated by our analyses of 24 large
federal agencies that considered the results of inspector general reports
and our reports published between July 2000 and September 2001. 3
The security challenges directly affect the ability to implement existing
laws and policies for protecting personal, proprietary, law enforcement,
and national security information. Such safeguards require the appropriate
tools to maintain confidentiality and ensure only authorized access,
sharing, and use. Without appropriate security tools, the protection of
this information will be at risk.
The information security challenges involved with a biometric system deal
with the protection of biometric data* whether they are a biometric watch
list or biometric reference templates stored in a central database or on a
token* and the transmission of those data. Table 11 gives examples of
operational issues, risks, and techniques related to binding individuals
to their biometric information when issuing them visas or passports with
biometrics. The binding process between a user and biometric information
is critical to the success of a biometric- based user- authentication
system. A process that does not have strong binding mechanisms will
provide little improvement over existing processes.
1 Government information security reform provisions of the FY 2001 Defense
Authorization Act* for example, 44 U. S. C. S:3534( a); Clinger- Cohen Act
of 1996* for example, 40 U. S. C.
S:11313( 6); Paperwork Reduction Act of 1995* for example, 44 U. S. C.
S:3506( g); and Computer Security Act of 1987* for example, 40 U. S. C.
S:11332.
2 U. S. General Accounting Office, Information Security: Opportunities for
Improved OMB Oversight of Agency Practices, GAO/ AIMD- 96- 110
(Washington, D. C.: September 24, 1996). 3 U. S. General Accounting
Office, Computer Security: Improvements Needed to Reduce Risk to Critical
Federal Operations and Assets, GAO- 02- 231T (Washington, D. C.: November
9, 2001).
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 103 GAO- 03- 174 Biometrics for Border Security
Table 11: Security Risks and Mitigating Techniques Event Adverse effect
Mitigating technique
Unauthorized changes are made to data in the central database* e. g., the
biometric data associated with S. Smith is changed so she can claim an
identity such as A. Smith.
The binding of a person to her biometric data is lost* in effect, she can
assume multiple identities.
Electronic signatures can ensure data integrity; system can periodically
check to ensure that the data and associated signature still agree.
Original data can be restored from a secure backup when a modification is
detected. a Unauthorized changes are made to a token*s data* e. g.,
biometric data originally stored with a given identity are replaced with
biometric data associated with an impostor.
The binding of a person to his or her biometric data is lost* in effect,
he or she can claim the identity of another person or assume multiple
identities, using the same token.
Electronic signatures generated by a central database at enrollment can
ensure a token*s data integrity. The data can be changed but changes would
be detected, since the system would not validate the electronic signature
generated during enrollment with the original data. A rogue government
official generates a false identity for a person with the correct
biometrics but altered name or birth date to bypass the system*s checks
for detecting suspicious individuals.
The binding of the person to her biometrics is not compromised, but the
system cannot ensure that travel documents are issued only to an
authorized person.
Split knowledge and dual control techniques can ensure that at least two
persons validate the identity data provided to the system. Also, once
identified, the electronic signature of the official who authorized the
token can easily be revoked. Biometric data on a token or in a database
are compromised by unauthorized disclosure.
Since the public may believe that biometric data are as confidential as a
Social Security number, their unauthorized disclosure may lead to identity
theft and a public relations problem.
A token*s biometric data can be encrypted to ensure that its loss or theft
does not compromise the data. Although encrypting the database might make
searching for duplicate values unrealistic, other controls can reasonably
limit access to biometric images to authorized persons and processes. a
Electronic signatures are commonly used to provide assurances that
unauthorized changes are not
made to data. They may also represent an individual or an entity. A
system- generated electronic signature should be (1) unique to the signer,
(2) under the signer*s sole control, (3) verifiable, and (4) linked to the
data in a way such that if the data are changed, the signature is
invalidated on verification. See U. S. General Accounting Office,
Information Security: Advances and Remaining Challenges to Adoption of
Public Key Infrastructure Technology, GAO- 01- 277 (Washington, D. C.:
February 26, 2001).
Source: GAO analysis.
Before any significant project investment is made, the benefit and cost
information should be analyzed and assessed in detail. A business case
should be developed that identifies the organizational needs for the
project. A clear statement of high- level system goals should drive the
overall concept of a U. S. border control system. Every aspect of the
overall system* from the selection of biometrics to the system
architecture* depends on the overall system goals. The high- level goals
should address the system*s expected outcomes* for example, Weighing Costs
and
Benefits
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 104 GAO- 03- 174 Biometrics for Border Security
binding a biometric feature to an identity (information such as name,
date of birth, place of birth) shown on a travel document,
identifying undesirable individuals on a watch list,
checking for duplicate enrollments,
verifying identities at the borders,
ensuring the adequacy of privacy protections, and
ensuring the security of the biometric information. Certain performance
parameters should also be carefully specified, including, among others,
the
time required to enroll people,
time required to verify each person*s identity by comparing the
biometrics against a stored template,;
acceptable overall FMR and FNMR, and
maximum population the system must handle. Similarly, not only must the
costs of the technology be considered but also the costs of the effects on
people and processes. A biometric- based border control system is bound to
require significant up- front investments and a certain level of recurring
costs to keep it operating. Weighed against these costs are the security
benefits that accrue from using the system. Analyzing this cost- benefit
trade- off is crucial when choosing specific biometrics- based border
control solutions. The consequences of performance issues* for example,
accuracy problems, their effect on processes and people, and their costs*
are also important in selecting a biometric technology.
The desired benefit of all the scenarios we describe is to prevent the
entry of travelers who are inadmissible into the United States. More
specifically, in both watch list scenarios, a biometric check could
improve security by adding a watch list check to the name- based watch
list checks already being performed. A biometric watch list could help
detect travelers who are trying to evade detection and who have
successfully established a separate name and identity. Biometrics that are
unique to these individuals
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 105 GAO- 03- 174 Biometrics for Border Security
should identify them in biometric checks against the entries in the watch
list. A biometric watch list could help detect certain travelers, even
when a name or other biographical information about an individual on a
watch list is unknown.
The quantitative benefit of the watch list scenarios (i. e., the number of
travelers prevented from obtaining U. S. travel documents or denied access
to the United States) would depend on the performance of the biometric
technology, the quality of the biometrics in the watch list, and the data
in the watch list. As we have described, the performance of the biometric
technology will determine the additional number of people apprehended as
well as the additional number of people identified incorrectly. The
performance of the biometric technology is also dependent on the size of
the biometric watch list. As more people are added to the watch list, the
probability of a false match for any given traveler increases. While
apprehending more people increases security, further questioning people
identified incorrectly increases the operational costs of implementing the
technology. The better the quality of the biometric in the watch list, the
more likely it is that the technology will correctly match a traveler to
it. Finally, if effective policies and procedures are not implemented to
populate the watch list, the system*s effectiveness will not be as great
as it could be.
For issuing passports and visas with biometrics, the key benefit is to
positively identify travelers as they enter the United States and to cut
down on the use of fraudulent travel documents, including counterfeit and
modified documents and impostors* use of legitimate documents. Travel
documents would continue to serve as evidence that the bearer has the
right of entry. The addition of biometrics can link the individual to the
travel document and serve as evidence that the present bearer of the
document is indeed the proper bearer. At ports of entry, INS inspectors
intercepted more than 114,000 fraudulent documents last year (see table
12). About one- third of the intercepted documents were U. S. passports or
visas.
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 106 GAO- 03- 174 Biometrics for Border Security
Table 12: The Number and Type of Fraudulent Documents INS Inspectors
Intercepted, Fiscal Year 2001
Document type Number intercepted
Border crossing cards 30,419 Alien registration cards 26,259 Nonimmigrant
visas 21,127 U. S. passport and citizenship documents 18,925 Foreign
passport and citizenship documents 15,994 Reentry permit and refugee
travel documents 702 Immigrant visas 597
Total 114,023
Source: INS.
The Census Bureau has estimated that between 7. 7 million and 8.8 million
unauthorized immigrants were in the United States in 2000. 4 INS has
estimated that the annual increase in the number of unauthorized
immigrants is about 275,000. 5 Of this number, INS estimates that about 60
percent of illegal immigration occurred *between the borders* and not at a
port of entry where people or documents could be inspected. INS estimates
that the remaining 40 percent of the undocumented population are
nonimmigrant overstays, meaning they entered legally on a temporary basis
but failed to depart. While it appears that current border control
processes reduced the annual number of unauthorized entrants by about one-
third, it is not known how many other travelers used fraudulent documents
to enter the United States. Today, inspectors check identity manually,
comparing photographs in a travel document with the face of the person
carrying the document.
Linking biometrics to visas and passports would help ensure that travelers
could not obtain travel documents under alternative identities once they
had already applied for initial documents and established a biometric
identity in the system. It would also help ensure that travelers who
crossed the borders were the persons depicted on their travel documents.
These two benefits could potentially decrease document fraud by making it
4 U. S. Bureau of the Census, Evaluating Components of International
Migration: Estimates of the Foreign- Born Population by Migrant Status in
2000, Population Division Working Paper 58 (Washington, D. C.: December
2001).
5 U. S. Immigration and Naturalization Service, Statistical Yearbook of
the Immigration and Naturalization Service, 2000 (Washington D. C.: U. S.
Government Printing Office, September 2002), 271* 74.
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 107 GAO- 03- 174 Biometrics for Border Security
harder to obtain a visa or passport under an assumed identity. The
scenario could also reduce the use of counterfeit visas and passports and
the use of legitimate documents by impostors.
Limitations to this approach are that a visa or passport biometric cannot
necessarily link a person to his or her true identity, although it can
bind him or her to a single identity within a system. A visa or passport
biometric system would make it more difficult for people to establish
multiple identities. Nevertheless, if the one identity a person claimed
were not his or her true identity, then the person would be linked to the
false identity in the biometric system.
Issuing visas with biometrics may have a limited effect because of the
relatively few travelers who must carry a visa to enter the United States.
While nonimmigrant aliens made 239 million border crossings last year,
many were not required to present a visa at the port of entry, including
Canadians, Mexicans who possessed a border crossing card, and aliens
entering through the visa waiver program. It is estimated that in only
about 22 million crossings were aliens required to have a visa to enter
the United States last year* about 15 million entered as visitors or with
task- specific visas (e. g., students), and another 7 million entered as
crew on airplanes or ships. Even though the current Mexican border
crossing cards are issued with two fingerprint templates on the card, it
is unclear how Mexicans would be affected by a decision to issue visas
with biometrics.
Issuing passports with biometrics might also have limited effect because
passports are not required of U. S. citizens when they enter the United
States from Canada or Mexico. While U. S. citizens made more than 179
million border crossings last year, it is not clear how many of them
needed or presented a passport to inspectors at the ports of entry.
While it is standard practice to quantify benefits in monetary terms, it
is difficult to do so for security applications. The monetary benefits of
keeping inadmissible people out of the country depend on the activities
undertaken while these travelers are in the country. Some inadmissible
people may simply affect the labor supply, while others may conduct
criminal or terrorist activities. Further information, including
behavioral assumptions, would be necessary in order to characterize the
value of preventing the entry of inadmissible persons.
As we have already stated, biometric technology is not a panacea for all
border security problems. For example, none of these scenarios addresses
two other key problems with border security. Previous INS estimates of
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 108 GAO- 03- 174 Biometrics for Border Security
illegal immigration were that about 60 percent of all illegal immigrants
entered *between the borders,* not at a port of entry where they could be
inspected. The scenarios we describe also will not help address problems
with aliens* overstaying their visits; aliens who overstay have already
presented themselves at a port of entry and were admitted by an inspector.
For each of the four scenarios, we created cost models to estimate the
cost of developing, implementing, and maintaining various biometric
processes. Besides including in the models the cost of purchasing the
biometric hardware, we estimated costs for additional hardware, software,
maintenance, personnel, training, and effects on other procedures in order
to derive life- cycle cost estimates. We used DOD*s definition of life-
cycle cost, which includes all costs the government incurs in designing,
developing, and operating a system through its life cycle, from its
initiation through disposal of the system at the end of its useful life.
We followed the cost element structure that DOD uses at acquisition
program milestone and decision reviews to assess major automated
information systems costs. Tailoring this structure to reflect our four
scenarios, we used it to standardize costs so that they could be compared
at a high level.
We present the costs in two parts. Initial costs represent the costs
required to plan, design, develop, and field the system. Recurring costs
represent the annual costs required to operate and continually maintain
the system to keep it in operation.
We estimated seven sets of initial cost elements: costs for systems
engineering and program management; development, installation, and
training; biometric hardware; biometric software; network infrastructure;
renovating consular facilities; and hardware infrastructure upgrades. We
estimated ten sets of recurring cost elements: program management;
biometric hardware maintenance; software and system maintenance; network
infrastructure maintenance; consular operating personnel; port of entry
operating personnel; communications; training; consular facility
maintenance; and annual supplies. (More details on the cost elements can
be found in appendix VI.)
We prepared the life- cycle cost estimates using fiscal year 2002 constant
dollars* that is, inflation was not considered for the multiple years over
which funds would be required for acquisition* and they represent rough
order of magnitude costs. In addition, the estimates in our technology
assessment are best guesses and should not be considered *budget
quality.* They attempt to provide a high- level view of what costs could
potentially be, given the assumptions we describe here. In order to System
Life- Cycle Costs
Assumptions
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 109 GAO- 03- 174 Biometrics for Border Security
develop budget- quality estimates, more details about the system to be
built are required, including an operational concept, detailed
requirements, site surveys, and vendor proposal data. Following are the
assumptions that frame the boundary of our cost estimates.
Scenario life- cycle cost estimates represent development and installation
time plus 10 years* operational life. Phasing of costs over time is
simplified, and actual schedules to both develop and install equipment and
infrastructure will most likely differ.
Biometric technologies* fingerprint, facial, and iris recognition*
represent standardization to a single vendor*s protocols. Biometric
technology costs represent the average costs of vendors* products. Four
flat fingerprints will be collected for fingerprint recognition.
There are 210 visa- issuing embassies and consulates worldwide. There are
4,500 passport acceptance offices. There are 3,950 primary and secondary
inspection stations at 400 ports of entry.
Personnel costs reflect both direct costs and indirect costs. Three
personnel will be needed to troubleshoot equipment at each port of entry,
or 1,200 additional staff.
No costs were estimated for
additional inspectors at ports of entry,
additional facility space for passport acceptance offices or at ports of
entry for primary and secondary inspections,
biometric equipment for exiting the United States, and
biometric security technology (e. g., encryption of biometric data). We
used the following assumptions to create the cost estimates for the two
biometric watch list scenarios:
The watch list database will include 10 million records.
Matches will be performed using facial recognition technology. Costs for
Scenarios 1 and
2: Watch List Checks
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 110 GAO- 03- 174 Biometrics for Border Security
To conduct watch list checks before issuing travel documents, facial
images will be generated by capturing the physical photographs applicants
present when they apply for a visa or passport.
The images will be collected and scanned at consulates and embassies for
visas and at passport acceptance offices and transmitted through
telecommunications resources to a central facility in metropolitan
Washington, D. C.
Estimates include costs for a primary central processing facility and a
contingency processing site. Table 13 summarizes the costs for the two
watch list scenarios.
Table 13: Estimated Costs for Watch List Checks Scenario Initial Recurring
1. Watch list check before issuing travel documents $52.8 $72.9 2. Watch
list check before entering the United States $330.2 $237.0
Note: Dollars are in millions. Source: GAO analysis.
In scenario 1, the major cost is additional consular staff to review
biometric watch list hits. It is assumed that each embassy or consulate
will require at least one additional foreign service officer to review
biometric watch list hits before visas are issued. If the performance of
the biometric technology requires more reviews and consequently more
staff, the cost of the scenario will increase. Of the $52. 8 million
initial cost, $33.1 million is for the placement of 221 additional foreign
service officers. Only $19.8 million is for the system*s development,
installation, and associated costs. Similarly, of the $72.9 million
recurring costs per year, $50.7 million is for additional foreign service
officers. Because it is unclear how many additional passport examiners
would be required to review biometric watch list hits for passports, we
have not included costs for additional passport examiners.
In scenario 2, adding facial recognition technology at the 400 ports of
entry greatly increases costs over scenario 1. The additional costs
related to developing and installing equipment at 3,950 primary and
secondary inspection stations at the ports of entry adds another $200
million to the system*s initial cost. (More details on the estimated costs
for conducting watch list checks with biometrics can be found in appendix
VI.)
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 111 GAO- 03- 174 Biometrics for Border Security
We used the following assumptions to estimate the costs of adding
biometrics to visas and to passports:
The number of visa applicants will remain constant at 10.3 million
annually. The number of travelers in the visa waiver program will remain
constant at 14 million annually.
The number of passport applicants will remain constant at 7 million
annually.
Enrolling travelers using a single biometric (whether for fingerprint,
facial, or iris recognition) is estimated at 6 minutes (10 applicants
enrolled per hour).
Enrolling travelers using multiple biometrics (e. g., fingerprint and
facial combined, fingerprint and iris combined, or fingerprint, facial,
and iris combined) is estimated at 10 minutes (6 applicants enrolled per
hour).
All current visa- issuing embassies and consulates and passport
acceptance offices will be equipped to collect biometrics from visa and
passport applicants, respectively.
Biometric token cards will be used to verify identities. We present cost
estimates for six different combinations of biometric technologies under
two different possibilities for issuing visas (see table 14). The State
Department receives about 10.3 million visa applications each year. In
fiscal year 2000, INS estimated that approximately 14 million individuals
traveled under the visa waiver program. If these travelers must obtain a
visa to travel to the United States, we assume that this same number would
also be required to have their biometric sample collected. An additional
14 million applicants increases the initial costs of the biometric system
by about 30 percent and annual recurring costs by about 50 percent. The
costs differ between the different combinations of biometrics because of
the different costs of the different types of equipment and the increased
time required to enroll people if more than one biometric is used. Costs
for Scenarios 3 and 4:
U. S. Visas and Passports with Biometrics
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 112 GAO- 03- 174 Biometrics for Border Security
Table 14: Estimated Costs for Issuing Visas with Biometrics Annual visa
applications 10.3 million with visa waiver program
24.3 million without visa waiver program Scenario 3: Issuing visas with
biometrics Initial Recurring Initial Recurring
Fingerprint recognition $1,422 $708 $1,879 $1,077 Iris recognition 1,419
707 1,876 1,075 Facial recognition 1,399 698 1,851 1,065 Fingerprint and
iris recognition 1,926 863 2,509 1,331 Fingerprint and facial recognition
1, 904 854 2,479 1,318 Fingerprint, iris, and facial recognition 2, 243
970 2,845 1,482
Note: Dollars are in millions. Source: GAO analysis.
Operating personnel and space at the embassies and consulates are a major
component of the cost estimates. Table 15 shows the initial and recurring
costs for consular operating personnel and space when using single
biometric and multiple biometrics. Depending on the combination of
biometric technologies, we estimate the costs of consular operating
personnel and space at 21 to 31 percent of initial costs and 23 to 29
percent of recurring costs. We did not include costs at ports of entry for
facility renovation or personnel to verify the biometrics of travelers
with visas as they enter the country.
Table 15: Estimated Consular Costs for Issuing Visas with Biometrics
Annual visa applicants 10.3 million 24.3 million Scenario 3: Issuing visas
with biometrics Initial Recurring Initial Recurring
Operating personnel Single biometric $75.9 $111.6 $114.9 $150.6 Multiple
biometrics 95.0 130.7 160.0 195.7 Space Single biometric 335.8 89.5 463.6
123.6
Multiple biometrics 378.2 100.9 563.7 150.5 Note: Dollars are in millions.
Source: GAO analysis.
Another major recurring cost is storage media for the biometric. At a cost
of about $15 per card for laser cards, this adds more than $150 million to
the recurring costs for 10.3 million visa applicants, or more than $360
million for 24.3 million applicants.
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 113 GAO- 03- 174 Biometrics for Border Security
Table 16 summarizes the costs of issuing passports with biometrics with
six different combinations of biometric technologies. Scenario 4 is by far
the most expensive. The primary cost difference between issuing visas with
biometrics and passports with biometrics is in the number of issuing
locations. While only 210 embassies and consulates can receive visa
applications, and even though fewer passport applications are received
annually than visa applications, there are more than 20 times more
passport acceptance offices (4,500) than embassies and consulates that
issue visas. This greater number of locations has a direct effect on the
estimates of initial and recurring costs for this scenario.
Table 16: Estimated Costs for Issuing Passports with Biometrics Scenario
4: Issuing passports with biometrics Initial Recurring
Fingerprint recognition $4,491 $1,574 Iris recognition 4,486 1,572 Facial
recognition 4,446 1,555 Fingerprint and iris recognition 6,694 1,978
Fingerprint and facial recognition 6, 655 1,961 Fingerprint, iris, and
facial recognition 8, 766 2,363
Note: Dollars are in millions. Source: GAO analysis.
As with the scenario in which visas are issued with biometrics, two major
costs are the costs of cards to store the biometrics and the personnel
required to collect the biometric sample from passport applicants. The
cost of cards adds more than $100 million per year to the costs of the
biometric system. While it is not clear how biometrics would be collected
at passport acceptance offices, we assumed that collection at each office
would require one additional staff person, for an additional annual cost
of $443.8 million. We did not include costs for additional space at the
offices.
In the scenarios where biometrics are added at the ports of entry (i. e.,
performing a watch list check before entering the United States and
issuing visas and passports with biometrics), the cost of additional
inspectors is not included. As we have previously described, the addition
of biometrics at the ports of entry would likely increase the inspection
time of each traveler. Without the addition of inspectors and the
corresponding space (i. e., inspection stations or lanes), delays would go
up at the ports of entry. We did not analyze how many additional
inspectors would be required to maintain current service times. These
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 114 GAO- 03- 174 Biometrics for Border Security
costs will have to be collected and analyzed as part of the preparation
for a budget- quality estimate should the government wish to pursue one of
these options. (More details on the estimated costs of issuing visas and
passports with biometrics can be found in appendix VI.)
Simulation is an analytical method meant to imitate a real- life system,
especially when other analyses are too mathematically complex or too
difficult to reproduce. Risk analysis uses both a spreadsheet model and
simulation to analyze the effect of varying the inputs to a modeled system
on the outputs. One type of spreadsheet simulation is Monte Carlo, which
randomly generates values for uncertain variables over and over to
simulate a model. The simulation results show not only the different
result values but also the probability (or certainty) of values.
We used both the initial and recurring costs as the forecast values and
ran the Monte Carlo simulation for each of the four scenarios. We applied
a probability distribution to each parameter that we thought could vary,
such as the costs for development and installation, annual operating
personnel, and additional square feet in embassy or consular facilities.
Table 17 shows our estimates, the level of certainty calculated by the
simulation for our estimates, and the cost for each scenario at the 90
percent certainty level. For issuing visas and passports with biometrics,
we simulated only two of the six possible combinations* one using a single
biometric and one using multiple biometrics.
Table 17: Cost Estimate Uncertainty Analysis for Four Scenarios Initial
Recurring Scenario Cost % certainty Cost at 90%
certainty Cost % certainty Cost at 90% certainty
1. Watch list check before issuing document $52.8 50% $53.3 $72.9 50%
$74.2 2. Watch list check and facial recognition 330.2 50 347.9 237.0 91
236.4 3. Visa
Fingerprint recognition 1, 879 70 1,923 1,077 91 1,059 Fingerprint and
facial recognition 2, 479 60 2,529 1,318 89 1,324 4. Passport
Facial recognition 4, 446 60 4,725 1,555 92 1,518 Fingerprint and iris
recognition 6, 694 70 6,892 1,978 91 1,953
Note: Dollars are in millions. Source: GAO analysis.
Developing, integrating, deploying, and maintaining biometrics to help
secure the nation*s borders will be costly. For example, the cost to
Uncertainty Analysis
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 115 GAO- 03- 174 Biometrics for Border Security
implement visas with biometric technologies would be on a par with a major
DOD weapons systems acquisition or FAA*s Standard Terminal Automation
Replacement System.
Underlying much discussion about the deployment of biometric technology
are questions about the sufficiency of information management laws, such
as the Privacy Act of 1974 and the Computer Security Act of 1987, to
protect civil liberties. 6 Periodic public surveys have revealed a
distinct unease with the potential ability of the federal government to
monitor individuals* movement and transactions.
The Privacy Act limits federal agencies* collection, use, and disclosure
of personal information. The act*s protections are keyed to the retrieval
of personal information by an *identifying number, symbol, or other
identifying particular assigned to the individual, such as a finger or
voice print, or a photograph.* 7 Accordingly, the Privacy Act generally
covers federal agency use of personal biometric information.
As a practical matter, however, the act is likely to have a more limited
application to biometric information in the context of border control.
First, it applies only to personal information regarding U. S. citizens
and lawfully admitted permanent resident aliens. 8 Second, the act
includes a number of exemptions that permit the disclosure of otherwise
covered information for internal agency use, for compatible *routine
uses,* and for law enforcement and national security purposes. 9
Representatives of civil liberties groups and privacy experts are
concerned about (1) the adequacy of protections for security, data
sharing, identity
6 The Privacy Act of 1974 (5 U. S. C. S:552a) and the Computer Security
Act of 1987, Public Law 100- 235 (15 U. S. C. S:278g- 3 and 4, 40 U. S. C.
S:11331, and 40 U. S. C. S:11332).
7 5 U. S. C. S:552a( a)( 4). 8 5 U. S. C. S:552a( a)( 2). 9 5 U. S. C.
S:552a( b), (j), (k). Effects on Privacy
and the Economy Privacy and Civil Liberties
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 116 GAO- 03- 174 Biometrics for Border Security
theft, and other identified uses and (2) secondary uses and *function
creep.* A significant number of concerns raised during our interviews and
conferences relate to the adequacy of protections under current law for
the large- scale data handling in a biometric system. Besides information
security, concern was voiced about an absence of clear criteria for
governing data sharing. The broad exemptions of the Privacy Act, for
example, provide no guidance on the extent of the appropriate uses law
enforcement may make of biometric information.
Of equal concern is a tendency for large organizations to develop
secondary uses of information; information collected for one purpose tends
over time to be used for other purposes as well. The history of the Social
Security number, for example, gives ample evidence of how an identifier
developed for one specific use has become a mainstay of identification for
many other purposes, governmental and nongovernmental. 10 Secondary uses
of the Social Security number have been a matter not of technical controls
but, rather, changing policy and administrative priorities. Further, some
are concerned that biometric information can potentially be linked to
multiple databases or to a vast national database. Questions being raised
include what data would be included or linked to a biometric
identification card; who would have access to such information,
legitimately or otherwise; and how people who can access such data could
use them.
Still others mention major concerns under the three headings of tracking,
profiling, and loss of anonymity. Tracking is real- time, or near- real-
time, surveillance in which a person*s movements are followed through her
biometrics- enabled transactions. While none of the scenarios we discuss
use biometric technologies for surveillance, we have heard concerns raised
about ways in which anonymity is likely to be undermined by surveillance.
For example, many civil liberties groups are extremely concerned about the
application of facial recognition technology for surveillance, which, like
video surveillance, could result in the loss of anonymity in public
places.
Profiling is the reconstruction of a person*s movements or transactions
over a specific period of time, usually to ascertain something about her
10 U. S. General Accounting Office, Social Security Numbers: Government
Benefits from SSN Use but Could Provide Better Safeguards, GAO- 02- 352
(Washington, D. C.: May 31, 2002).
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 117 GAO- 03- 174 Biometrics for Border Security
habits, tastes, or predilections. Profiling for race, ethnicity, or
national origin has caused much public debate in recent years. Tracking
and profiling can destroy anonymity. The lack of clear policy goals and
any flaws in the operation of biometric technology could compound all
these concerns.
Concerns have also been raised about whether certain biometric data might
reveal medical predispositions or personal health histories whose use
could result in denial of insurance coverage or employment. For example,
while not currently viable, the use of DNA matching as a biometric
technology would be of concern because of the personal medical information
that could be gleaned from it.
Not only are there concerns with secondary uses, but there are also
concerns with unauthorized uses. Our recent studies on identity theft and
studies by others reflect the difficulty of accurately measuring
identification theft. 11 Developing a large- scale interoperable system
such as a watch list system or issuing visas or passports with biometrics
could increase the risks of identity theft and other unauthorized uses of
personal information if the biometric data are not properly protected.
Biometrics industry groups, while expressing their appreciation of privacy
concerns, have responded by saying that biometric products are *privacy
neutral* and that it is how they are used that reflects either privacy
invasion or privacy protection. IBG has developed a framework for defining
the potential privacy risks borne by specific biometric technologies and
their deployment. IBIA is also advocating on behalf of the industry to
create responsible use guidelines and public policy. Industry groups
emphasize self- regulation, which some privacy groups assert is not enough
because markets are erratic and because, they say, the high value placed
on data means incentives for violation are too high. Nonindustry groups
have also developed privacy frameworks. The Internal Revenue Service (IRS)
published a guide to developing privacy impact assessments for information
technology. Also, the RAND Corporation developed a four- step approach for
responding to sociocultural concerns about biometrics. Table 18 combines
some of the salient characteristics of the guidelines IBG, IRS, and RAND
developed and outlines many of the
11 U. S. General Accounting Office, Identity Fraud: Prevalence and Links
to Alien Illegal Activities, GAO- 02- 830T (Washington, D. C.: June 25,
2002).
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 118 GAO- 03- 174 Biometrics for Border Security
questions to be answered in assessing the potential effect on privacy from
any new biometric system.
Table 18: Summary of Biometric Systems Privacy Guidelines Issue Guideline
Scope and capabilities Does the system have a clearly and narrowly defined
purpose? Who will use the system? Have potential system capabilities been
evaluated? Has there been an evaluation of a range of alternative choices,
including biometrics? What types of information will be available through
the biometric? Will the biometric information be used as a universal
unique identifier? Will the storage of biometric information include
extraneous information? Will the system store the original biometric data?
Data protection Will the system separate biometric information from other
types of personal information?
What procedures will limit access to the system? Who will have access? Do
other systems or agencies share or have access to data in this system? If
data are being consolidated, what controls protect them from unauthorized
access or use? How will the system ensure accuracy? What are the sources
of information in the system? How will data collected from other sources
be verified for accuracy? Will the system derive new data or create
previously unavailable data about an individual through aggregation from
the information collected? Can the system make determinations about
individuals that would not be possible without the new data? How long will
data be retained in the system? What are the procedures for eliminating
the data at the end of the retention period? While the data are retained
in the system, what are the requirements for determining whether the data
are sufficiently accurate, timely, and complete to ensure fairness? User
protection Will users have the ability to unenroll?
Will users be able to access and correct their biometric information? Will
there be procedures for anonymous enrollment? Will the system and its use
ensure individuals* equitable treatment? If the system is operated in more
than one site, how will consistent use of the system and data be
maintained? Will the system be able to identify, locate, and monitor
individuals or groups of people? What controls will prevent unauthorized
monitoring? Disclosure, auditing, accountability, and oversight Will there
be full disclosure of audit data?
Will the system*s purpose be disclosed? Will the enrollment, verification,
and identification processes be disclosed? Will the names of the
individuals and entities responsible for system operation and oversight be
accessible? Will users be informed about optional versus mandatory
enrollment? Will a board or committee assess biometric data policies? Will
third parties oversee the system? Who will review requests for access to
biometric data? Who ensures that the biometric program is responsive to
privacy concerns?
Source: GAO summary of data from IBG, IRS, and RAND Corporation.
These guidelines can help decision makers and other stakeholders approach
privacy issues and determine the appropriate balance of privacy and
security to build into the system. However, because there is no general
agreement on the answers to these guidance questions, further policy
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 119 GAO- 03- 174 Biometrics for Border Security
decisions are needed. The range of unresolved policy issues suggests that
questions surrounding the use of biometric technology center as much on
management policies as on technical issues.
As previously described, implementing biometrics could lengthen the
process of obtaining travel documents or entering the United States. At
some posts, visas are issued the same day applications are received. If
significant time is required to resolve biometric watch list or visa
database hits, issuing visas could be delayed. At the ports of entry, in
order to avoid long lines of pedestrians and vehicles, each inspection has
to be fast* according to INS officials in El Paso, Texas, for example, any
time longer than 15 seconds would cause staggering delays. Even so, the
busiest ports of entry regularly have delays of 2 to 3 hours.
Checking the biometric identity of passengers in vehicles is especially
challenging. In a biometric system, would passengers have to exit their
vehicles in order to have their biometrics checked? Any increase in
inspection times could compound delays. Delays inconvenience travelers and
increase their costs. Studies have been conducted on the value of travel
time, and further studies in this area could help determine whether the
increased security could result in fewer visits to the United States or
lost business to the nation.
While biometrics- based border control would affect regional economies and
various economic sectors, it is difficult to quantify its effect. However,
we can postulate that the travel and tourism industry might be adversely
affected. Spending by international travelers in the United States totaled
about $103 billion in 2000 and $90 billion in 2001. This spending is
particularly important for California, Florida, and New York, which
together account for more than half of all spending by international
overseas visitors. If a biometric system made it more difficult to obtain
visas for whatever reason, from higher visa fees to longer time between
application and issuance, international travelers might choose to visit
other countries instead. Further, there are concerns that if fingerprint
recognition technology were used, the number of visitors from countries
such as Japan would decrease dramatically because of those societies*
aversion to fingerprinting.
At the regional level, biometrics could significantly affect trade with
Canada and Mexico, the nation*s largest trade partners, with total trade
amounting to $653 billion in 2000. More than 80 percent of all Canadian
exports are destined for the United States. If biometric identification
Convenience for Travelers
Economic Impact
Chapter 5: Applying Biometrics to Border Control: Challenges and
Implications
Page 120 GAO- 03- 174 Biometrics for Border Security
checks increased waiting time at land crossings, local merchants on both
sides of the border could lose sales. Biometric systems might also have a
profound effect on Mexico*s maquiladora industry* the most dynamic sector
of the Mexican economy, adding 1,400 new plants and 640,000 new jobs since
the 1994 implementation of the North American Free Trade Agreement
(NAFTA). 12 U. S.* NAFTA partner trade is concentrated at a few ports. In
2000, 10 accounted for 73 percent of all North American trade by land.
Biometrics- based border control would have to be implemented carefully at
these ports to preserve the flow of trade.
The use of biometrics in a border control system in the United States
could affect the number of international visitors and how other countries
treat visitors from the United States. Much visa issuance policy is based
on reciprocity* that is, the process for allowing a country*s citizens to
enter the United States would be similar to the process followed by that
country when U. S. citizens travel there. If the United States requires
biometric identifiers when aliens apply for a visa, other nations may
require U. S. citizens to submit a biometric when applying for a visa to
visit their countries. Similarly, if the United States requires other
countries to collect biometrics from their citizens and store the data
with their passport for verification when they travel here, they may
require the United States to place a biometric in its passports as well.
As more countries require the use of biometrics to cross their borders,
there is a potential for different biometrics to be required for entering
different countries or for the growth of multiple databases of biometrics.
Unless all countries agree on standard biometrics and standard document
formats, a host of biometric scanners might be required at U. S. and other
ports of entry. ICAO plans to standardize biometric technology for
machine- readable travel documents, but biometric data- sharing
arrangements between this country and others would also be required.
12 Maquiladora refers to a Mexican company that imports, on a duty- free
basis, machinery, equipment, and materials for the manufacture of finished
goods for subsequent export. International Relations
Chapter 6: Summary Page 121 GAO- 03- 174 Biometrics for Border Security
In this report, we have considered a number of leading and emerging
biometric technologies that could potentially be used for securing the
nation*s borders. The seven leading biometric technologies include facial
recognition, fingerprint recognition, hand geometry, iris recognition,
retina recognition, signature recognition, and speaker recognition. Among
the emerging technologies, we considered vein scan, facial thermography,
DNA matching, odor sensing, blood pulse measurement, skin pattern
recognition, nailbed identification, gait recognition, and ear shape
recognition. Our assessment is based on a snapshot of biometric
technologies as they existed in early 2002.
Of the seven leading technologies, fingerprint recognition, facial
recognition, iris recognition, and hand geometry appeared to be suitable
for border security. These technologies could be used to associate a
person*s identity with travel documents and thus deter fraudulent use of
travel documents. Some of these technologies could be used to check to see
whether a person is on a watch list. Of the four technologies, hand
geometry is not very good at identifying one person in millions and,
therefore, is not suitable if we want to search the biometric database to
determine whether a person has previously enrolled in the database.
However, hand geometry can be used to verify identity in combination with
another technology. We found that the emerging biometric technologies are
in various stages of development and have not yet been used in border
control applications.
When it comes to effectiveness, all biometric technologies share a common
characteristic. Every time a biometric feature is captured, it is always
slightly different from the feature that was originally captured and
stored in the system. Also, sometimes the biometric device cannot capture
the biometric feature at all. Thus, all biometric technologies suffer from
three types of error* the failure to capture a biometric feature, falsely
not matching a biometric even though the person*s biometric is in the
system, or falsely matching a biometric. Each biometric technology has
different levels of these errors, and the errors depend on many different
factors, including the operational environment and security- level
setting. For example, it is possible to trade off the false match and
false nonmatch errors against each other. Thus, the effectiveness of a
biometric technology depends on how it is used in an overall system.
Chapter 6: Summary
Chapter 6: Summary Page 122 GAO- 03- 174 Biometrics for Border Security
It is important to recognize that biometric technology would be but one
component of the decision support systems that determine who is allowed to
enter the United States and who is not. As we have described, these
decisions are generally two- step processes. First, a decision must be
made to determine whether or not to issue a traveler a U. S. travel
document. Second, a decision is made at a port of entry on whether to
admit the traveler into the country. While the first step is not always
executed, depending on the nationality of the traveler, all legal entries
into the United States must be through an official port of entry.
The task of sorting admissible travelers from inadmissible ones is now
conducted by using information systems for checking names against watch
lists and by using human, manual recognition capabilities to determine
whether a photograph on a travel or identification document matches the
person who seeks entry into the United States. The introduction of
biometrics into this process could help automate the identification of
travelers.
In this report, we explored the use of biometrics in two types of systems.
In one system, biometrics can check a person*s face against a watch list
of facial images and provide alerts if there is a potential match. In
another system, the identity of travelers can be verified by comparing
their proffered biometrics (e. g., a fingerprint) against stored templates
that are associated with their travel documents.
We have found that three key considerations must be addressed before a
decision is made to design, develop, and implement biometric technologies
in a border control system:
1. Decisions must be made on how the technology will be used. 2. A
detailed cost- benefit analysis must be conducted to determine that
the benefits gained from a system outweigh the costs. 3. A trade- off
analysis must be conducted between the increased
security, which the use of biometrics would provide, and the effect on
areas such as privacy and the economy.
As we have described, technology and people each have a role in executing
processes to achieve a goal. Before anything else can be defined, the
high- level goals of a system with biometrics must be clearly articulated.
System goals are based on business or public policy needs. For example, a
goal could be to prevent known inadmissible people from Key Considerations
in Using Biometrics for Border Control
Chapter 6: Summary Page 123 GAO- 03- 174 Biometrics for Border Security
entering the country. Based on the high- level goals, a concept of
operations can be developed that embodies the people, process, and
technologies to achieve the goals. To put together the concept of
operations, a number of inputs have to be considered, including legal
requirements, existing processes and infrastructure used, and known
technology limitations. Performance requirements should also be included
in the concept of operations. For example, an average inspection time or
an average time to issue a visa could be included as performance
requirements. Any process reengineering that would be required to
accommodate the new technology should also be handled during this stage.
For a biometric system, this could include new processes to conduct
inspection of passengers in vehicles and to maintain the database of
biometric reference templates.
Once the concept of operations is complete, the requirements of the
biometric technology system can be developed and a particular technology
solution can be selected. The system requirements are derived from the
role of the technology system as defined in the concept of operations.
Detailed system requirements should include functional and performance
parameters, interface requirements, usability requirements, system quality
requirements, and security and privacy issues.
The primary question to be asked when selecting the technology solution is
whether it can support the requirements specified in the concept of
operations and the system requirements. Particular attention must be paid
to error measures, such as the false match rate, false nonmatch rate, and
the failure to enroll rate. Concerns about the distinctiveness and
stability of the technology, as well as its adherence to industry
standards, should also be addressed. Because biometric technologies have
not been used in applications as large as border control, further research
may be required to establish the distinctiveness and stability of the
biometric features. Distinctiveness has two aspects* how distinct a
biometric feature is across a population and how many different biometric
features are needed to uniquely identify an individual in a given
population. Stability refers to how the biometric features change as a
person ages. It is unclear whether a biometric captured during enrollment
will still be properly matched with an acceptable level of accuracy after
5 years, for example.
Chapter 6: Summary Page 124 GAO- 03- 174 Biometrics for Border Security
A cost- benefit analysis must be conducted to justify the investment in a
biometric border control system. 1 The benefits gained from a biometric
border control system should be based on how well the system achieves the
high- level goals. For example, if a goal of the system is to prevent
known inadmissible people from entering the country, one of the benefits
should be based on an estimate of how many additional inadmissible people
are intercepted or deterred from entering the country with this system.
The performance of the biometric system measured by its error rates would
directly affect the expected benefit.
All life- cycle costs for the biometric system should be included in the
analysis. The costs of the system include design, development,
implementation, operations, and maintenance costs. Costs associated with
the new business processes needed to use the new system should also be
included, such as the costs of personnel to enroll people in a biometric
system and the office space required to conduct the operation. Additional
people and processes, and their costs, required as a result of performance
limitations of the technology also must be included.
Finally, a trade- off analysis must be conducted between increasing
security and its effect on areas that are harder to quantify, such as
privacy, convenience, and the economy. Even if the cost- benefit analysis
shows that the benefits outweigh the costs, the effect of increasing
security may affect these areas to such a degree that the biometric system
should not be undertaken.
Complying with the legal requirements for privacy is necessary to
implement the system. Further, a system that does not include adequate
protections for privacy may encounter barriers from users, who may not
accept it during operation. The historical trade- off in any security
system is between security and convenience. If a security system is not
easy to use, people will stop using it. Similarly, if the process to enter
the United States becomes too hard or time- consuming, people may choose
to stop coming. This effect may manifest itself as an economic impact on
the country as retail and trade diminish. Finally, the effect on the
nation*s dealings with other countries and their citizens must also be
considered. International travel involves not only U. S. citizens but also
citizens from
1 For more information on cost- benefit analysis, see Office of Management
and Budget,
Guidelines and Discount Rates for Benefit- Cost Analysis of Federal
Programs, Circular A- 94 (Washington, D. C.: October 29, 1992; rev.
January 22, 2002).
Chapter 6: Summary Page 125 GAO- 03- 174 Biometrics for Border Security
other countries. Hardships imposed on those citizens may result in
reciprocal procedures being imposed on our citizens.
We defined four different scenarios in which biometrics could be used in a
border control system. We considered each of the key considerations for
the different scenarios. We did not answer the questions of whether each
scenario is cost- beneficial or whether the gains in security outweigh the
effects on privacy and the economy, but we did describe some of the
effects that introducing biometrics into each scenario would have on
people and processes, the benefits to be gained from the system, and the
limitations of biometric technologies. We also described the effect of
these limitations on the benefits and the effects on privacy, the economy,
and international relations.
In addition, we prepared rough order of magnitude costs for each scenario.
As summarized in table 19, initial investments could range anywhere from
less than a hundred million dollars for a watch list application to
billions for biometrics- enabled passports. Many of the recurring costs
would be for the salaries of personnel required to enroll people and
operate and maintain the system.
Table 19: Estimated Costs for Implementing Border Security Scenarios
Scenario Initial cost Annual
recurring cost
1, Watch list check before issuing travel documents $53 $73 2. Watch list
check before entering the United States 330 237 3. Issuing visas with
biometrics 1,399* 2,845 698* 1,482 4. Issuing passports with biometrics
4,446* 8,766 1, 555* 2,363
Note: Dollars are in million. Source: GAO analysis.
These costs have to be weighed against the benefits, which include
reducing the fraudulent use of travel documents and automating the
determination of whether travelers are on a watch list as they arrive at a
port of entry. By binding an individual*s biometric features to a travel
document* either by storing the features on a token, such as a smart card
the traveler carries, or by associating the identity with the biometric in
a central database* the border inspection process would allow travelers to
enter only if the stored biometric matches the biometric the individual
presents at inspection. In a typical watch list, photographs of
undesirable High- Level Analysis of
Four Scenarios Using Biometrics
Chapter 6: Summary Page 126 GAO- 03- 174 Biometrics for Border Security
people would be in a watch list of facial images, and a facial recognition
system would automatically compare the facial image of each traveler with
that in the watch list and identify potential matches. Inspectors can then
further investigate those matches.
While using biometric watch lists and incorporating biometrics into travel
documents could improve border security, the use of biometrics alone will
not prevent the illegal entry of foreign terrorists and others into the
country. For example, biometrics cannot prevent the illegal entry of those
who do not enter through official ports of entry. Further, even at the
legal ports of entry, unless all travelers are required to have their
biometrics checked, it is possible that a traveler could bypass the
biometric check. For example, if U. S. citizens are not required to enroll
their biometrics to travel internationally and an alien could convince the
inspector that he or she is a U. S. citizen, he or she could avoid the
biometric check.
The use of biometric technologies could also have a significant effect on
many different areas, from privacy concerns to the economy. While it
appears that the Privacy Act of 1974 generally covers federal agency use
of personal biometric information, as a practical matter the act is likely
to have a more limited application in the context of border control
because the act includes exemptions for law enforcement and national
security purposes and does not cover nonimmigrant aliens. Civil liberties
groups and privacy experts expressed concern about the adequacy of
protections under current law for the large- scale data handling in a
biometric system. Besides information security, concern was voiced about
an absence of clear criteria for governing data sharing. Another concern
was raised about the potential for secondary uses of biometric data and
what other data would be linked to a biometric identification; who would
have access to such information, legitimately or otherwise; and how people
who can access such data could use them.
When used in border control, biometric technologies also affect the
economy and international relations. We can postulate that the travel and
tourism industry might be adversely affected, because biometrics- enabled
visas may take longer to issue and may be considered more trouble than
they are worth. Spending by international travelers in the United States
totaled about $103 billion in 2000. At the regional level, biometrics
could significantly affect trade with Canada and Mexico, the nation*s
largest trade partners, with total trade amounting to $653 billion in
2000. If biometric identification checks result in increased waiting times
at land crossings, local merchants on both sides of the border could lose
sales.
Chapter 6: Summary Page 127 GAO- 03- 174 Biometrics for Border Security
Using biometrics in a border control system in the United States could
affect how other countries treat visitors from the United States. The
reciprocity of visa issuance policy implies that if the United States
requires biometric identifiers when aliens apply for a visa, other nations
may require U. S. citizens to submit a biometric when applying for a visa
to visit their countries. Similarly, if the United States requires other
countries to collect biometrics from their citizens and store the data
with their passports for verification when they travel to the United
States, these countries may require the United States to place a biometric
in its passports as well. As more countries require the use of biometrics
to cross their borders, there is a potential for different biometrics to
be required for entering different countries or for the growth of multiple
databases of biometrics. Unless all countries agree on standard biometrics
and standard document formats, a host of biometric scanners might be
required at U. S. and other ports of entry.
To address the role of biometrics in the overall border security problem
and the high- level goals that can be achieved by using biometrics, a
riskbased approach could be used. As we have previously reported, risk
management is the foundation of effective security. 2 The approach to good
security is fundamentally similar, regardless of the assets being
protected, whether information systems security, building security, or
homeland security. The answers to five basic questions can help determine
the role of biometrics in a border security solution:
What am I protecting?
Who are my adversaries?
How am I vulnerable?
What are my priorities?
What can I do? A decision to implement one or more of the scenarios we
have defined and analyzed in this report should be founded on a risk-
based approach that
2 U. S. General Accounting Office, National Preparedness: Technologies to
Secure Federal Buildings, GAO- 02- 687T (Washington, D. C.: April 25,
2002). The Role of
Biometrics in Border Security
Chapter 6: Summary Page 128 GAO- 03- 174 Biometrics for Border Security
answers these questions. These scenarios are by no means the only ways to
implement biometrics to assist in the border control mission. Some of
these scenarios could be implemented partially, or brand new scenarios
could be used. For example, it could be possible to check biometrically
enabled travel documents only at air ports of entry instead of at all
ports of entry. Some have suggested that only selected travelers*
biometrics be checked at the ports of entry instead of all travelers*
biometrics. The selection could be random or based on travelers who fit a
particular description. Another possible implementation of biometrics is
within optional programs similar to INSPASS, in which travelers
voluntarily choose to have their biometrics enrolled. In such a system,
travelers would enroll with the expected benefit that they would be able
to cross into the United States more quickly. Regardless of the system
concept, decisions must be made that determine how the biometric
technology will be used, if the benefits outweigh the costs, and what the
effects on areas such as privacy and the economy will be. While a partial
implementation may be less costly, the security benefits gained from such
a system may also be less.
We have also noted that biometric technologies are not a panacea for the
border security problem. It is important to realize that even with
biometrics, many system dependencies cannot be controlled wholly by a
technological solution. For example, if biometrics are included with
visas, the process will still require establishing a traveler*s initial
identity in the biometric system. Once that identity is established, the
benefits of strongly binding the individual to that identity can be
gained. However, the system depends on the process used initially to
establish that identity* that is, the applicant*s presentation of a
passport from his or her country. If the foreign country does not have
adequate controls over the way it issues passports or, worse, deliberately
issues passports with false identities, an individual could obtain a U. S.
visa with a biometric unless additional processes are in place to further
verify the applicant*s identity. These processes are not a part of the
biometric system but are still important for border security.
The population of a biometrics- based watch list is also dependent on
nontechnological processes. As we have previously stated, the policies and
procedures governing the population of a biometric watch list are critical
to the success of the program. The success of the program depends on the
effectiveness of the law enforcement and intelligence community to
identify individuals to place on the watch list. People who are not on the
watch list cannot be intercepted when trying to obtain a travel document
or entering the country. Further, biometrics cannot help in detecting
Chapter 6: Summary Page 129 GAO- 03- 174 Biometrics for Border Security
illegal entry into the United States through other than the official ports
of entry. They also cannot help in detecting aliens who enter the country
legally but then overstay the terms of their visit.
Using a risk- based approach should help in the development of a biometric
system*s high- level goals and its concept of operations. The answers will
help point out the system*s limitations and what it will not be able to
provide. They could also play a role in determining the appropriate
balance between increasing security and cost and operational
considerations as well as the effect on issues such as privacy and the
economy. With these answers, the proper role of biometric technologies in
border security can be determined.
We provided a draft of this report to the Department of State and the
Department of Justice for their review.
In written comments on a draft of this report, the Department of State
stated that it appreciated the thorough and balanced approach we took in
our assessment of the use of biometrics for border security. State found
the overall thrust of the report to be in keeping with its own
considerations of how biometrics could be used in admitting individuals to
the United States and how it could be integrated into the existing process
for visa and passport applications. State agreed with us on the need for
high- level policy decisions such as defining the specific uses of
biometric data and performing a cost- benefit analysis that weighs the
effectiveness and security benefits of biometrics against costs and the
probable implications or consequences of implementation, including
economic, civil liberty, and foreign policy concerns. State believed that
policy decisions must be made before the successful selection, execution,
and implementation of a border security program involving biometrics.
State noted that it is developing additional options for the
implementation of a biometric program whose final estimated costs may
differ. State also provided written technical comments on the draft, which
we incorporated as appropriate.
In written comments on a draft of this report, the Department of Justice
expressed concern that the report did not (1) properly consider an overall
border security strategy; (2) adequately recognize the draft NIST
certified Agency Comments
and Our Evaluation Department of State
Department of Justice
Chapter 6: Summary Page 130 GAO- 03- 174 Biometrics for Border Security
standards recommendations for biometrics, tamper- resistant travel
documents, or interoperability; or (3) fully explore the advantages of
some biometrics over others. Justice also said that the draft contained
analytical weaknesses related to a misunderstanding of the false match
rate metric and to performance data and levels.
First, on the subject of an overall border security strategy, Justice
explained that it has prepared such a strategy and that the U. S.
government is continuing to consider this strategy. Justice believed that
the implementation of this strategy would require major improvement in
border systems. Further, according to Justice, if the use of biometrics
were limited to visa applicants, who cover only about 3 percent of
visitors to the United States, the impact on preventing the entry of
potential terrorists into the country would be marginalized.
We requested a copy of the strategy from Justice on October 11, 2002, but
as of October 24, 2002, we had not received the strategy document from
Justice. While we did not have the opportunity to review Justice*s
strategy document, we do agree with Justice*s assertion that an overall
border security strategy is needed. Concerning Justice*s point that visa
applicants comprise only 3 percent of visitors to the United States, it is
pertinent to note that limiting the use of biometrics to visa applicants
would still target individuals living among the countries that are a
higher risk of directing terrorism at the United States. Whether the use
of biometrics should be limited to visa applicants should be based on
Justice*s border security strategy. We have previously stressed the need
to develop and implement a homeland security strategy in coordination with
all relevant partners. 3 This strategy should be comprehensive and should
encompass steps designed to reduce our vulnerabilities, deter attacks,
manage the effects of any successful attacks, and provide for appropriate
response. The strategy should involve all levels of government, the
private sector, individual citizens both here and abroad, and other
nations. The strategy should also use a risk management approach to focus
finite national resources on areas of greatest need. In this report, we
reiterate the need for a risk- based strategy for the use of biometric
technology in border security.
3 U. S. General Accounting Office, Homeland Security: Challenges and
Strategies in Addressing Short- and Long- Term National Needs, GAO- 02-
160T (Washington, D. C.: November 7, 2001).
Chapter 6: Summary Page 131 GAO- 03- 174 Biometrics for Border Security
Second, Justice was concerned that this report presents information about
biometrics that is inconsistent with the results of a NIST study required
by the USA PATRIOT Act. In particular, Justice stated that the intended
application must (1) employ a biometric that is able to establish and
verify a unique identity in a population of hundreds of millions, (2) be
used to run a check against criminal records, and (3) operate with a very
low risk of false positive reads and with a verification process that is
not rendered ineffective in different border, lighting, and weather
conditions.
The results of the NIST study were not available during our technology
assessment. However, we provided a draft of this report to NIST and
received comments from NIST reviewers, which we incorporated, where
appropriate, into the report. Further, we do not see an inconsistency
between our position and Justice*s description of the NIST study results.
We consider these to be examples of items that would be defined in a
concept of operations or a system requirements specification. The thrust
of our report has been to point out the possibilities and not to select a
specific biometric for border security* primarily because the selection
comes after the concept of operations and requirements is developed. The
requirements Justice described are what the department, with NIST*s
assistance, is defining as the requirements for a biometrics border
control system.
Third, Justice stated that there are certain advantages to using
fingerprints over other biometrics. For example, Justice cited the
requirement in the Immigration and Nationality Act for aliens to be
registered and fingerprinted, unless waived at the discretion of the
Secretary of State. Justice further cited the law enforcement value of
using fingerprint recognition for biometric identification on a large
scale. Justice stated that unlike other biometric data, fingerprints are
left at crime scenes. Further, Justice stated that we do not consider that
the use of fingerprints would allow for a search against records stored in
IAFIS to check for criminal history.
We acknowledge the qualities of fingerprint recognition raised by Justice.
However, as we have described, the additional benefits Justice described
should be included in the cost- benefit analysis that weighs the security
benefits gained from a biometrics border control system against the costs
of building the system. A benefits assessment should be based on how well
the system achieves the high- level goals defined for the system. For
example, if the ability to collect biometric samples at crime scenes is a
requirement, it should be factored into the goals and requirements
definition of the system. An evaluation of the technologies against the
Chapter 6: Summary Page 132 GAO- 03- 174 Biometrics for Border Security
requirements would then show that fingerprint recognition is the only
technology that can meet that particular requirement.
Justice also believed that the draft report contained analytical
weaknesses related to an insufficient analysis of large systems, misuse of
performance metrics, and the reporting of performance data. Specifically,
Justice stated that the draft did not provide sufficient analysis of large
systems and did not define and use the false match rate metric correctly.
Justice pointed out that the report provided incorrect performance data on
IAFIS. Justice further cited the problem of having to manually resolve
false matches and how the size of the database affects the number of false
matches.
In the report, we state that none of the biometric technologies have been
used with databases containing hundreds of millions of individuals. We
have clarified the definition of false match rate and, on the basis of
written technical comments provided by Justice, we have incorporated the
performance data on IAFIS. We acknowledge Justice*s point of having to
manually resolve false matches. We state in the report that the
performance of the biometric technology and its effect on people and
processes are important in the selection of the technology. We also
describe the potential effects of poorly performing biometrics on the
border control process.
Finally, Justice stated that the *draft report infers that any move toward
biometrics be made slowly and cautiously.* Justice agreed that it is
important to proceed judiciously but pointed out the sense of
congressional urgency raised in the USA PATRIOT Act and the Enhanced
Border Security and Visa Entry Reform Act.
We appreciate the urgency in developing a biometric system for border
security, but a timely decision to invest in such a system should still be
made in accordance with applicable federal regulations and best practices
for acquiring information technology systems.
Justice also provided written technical comments on the draft, which we
incorporated as appropriate. We have included Justice*s comments in their
entirety in appendix VIII.
Chapter 6: Summary Page 133 GAO- 03- 174 Biometrics for Border Security
We provided a draft of this report to 16 different organizations for their
review. Individuals from these organizations were selected because of
their assistance during the data collection phase of our work. In
addition, several of them attended one of our two biometric and border
security meetings, convened for us by the National Academy of Sciences.
The reviewers represented government, industry, and academia. We received
comments and suggestions from 10 reviewers. The comments ranged from
correcting technical inaccuracies to highlighting certain aspects of the
assessment that the reviewers considered important.
Several reviewers commended us for putting together such a thorough report
in a short time. One reviewer said that the report contains a wealth of
information that will be useful to all biometric practitioners and
researchers. Another reviewer noted that we have been able to successfully
develop a cohesive report despite the difficulties associated with the
wide variety of information available from vendors and studies. A reviewer
felt that this will be a milestone report* head and shoulders above any
other report on biometrics for border security.
In their comments, several reviewers cited their agreement with specific
findings in our report, particularly with the limitations of biometric
technologies in helping to secure the nation*s borders. Specifically,
reviewers agreed that biometric technologies are not a panacea for the
border security problem and that they are just one component of the
decision support systems that determine who is allowed to enter the United
States and who is not. Further, reviewers agreed that biometrics cannot
necessarily link a person to his or her true identity, although it can
bind an individual to a single identity within a system. One reviewer
concurred with the report*s point about the difficulty in quantifying the
benefits of security improvements.
Regarding the accuracy of a biometrics system, reviewers were concerned
that we had not clearly defined and used FMR and FNMR as performance
metrics. Reviewers also mentioned that we should not always equate FMR to
false accept rate and FNMR to false reject rate. Similarly, reviewers were
concerned that we did not separate out the results of larger tests
involving many enrolled individuals from smaller ones and that in
tabulating the performance of the biometric technologies, we had mixed
results from verification and identification into the same table.
On the basis of these comments, we clarified the definitions of FMR and
FNMR, highlighted how the same technology can have different performance
requirements for verification and identification, and selected External
Reviewers*
Comments
Chapter 6: Summary Page 134 GAO- 03- 174 Biometrics for Border Security
comparable test results when summarizing the performances of the biometric
technologies in a table.
Reviewers commented on our conclusion that hand geometry was not suitable
for border control, pointing to its current use in border control
applications. We had ruled out hand geometry for border control because it
is not distinctive enough to perform identification matches when checking
watch lists or for duplicate enrollment. Reviewers explained that for
issuing visas or passports with biometrics, a more distinctive biometric
such as fingerprint could be used for the duplicate enrollment check,
whereas a simpler biometric such as hand geometry could be used when
performing one- to- one verification at ports of entry. We incorporated
this idea into our report so that hand geometry is now listed as a viable
technology for border security.
On the subject of biometric standards, reviewers commented that in order
to avoid being tied to a vendor*s proprietary biometric template format,
any biometric system has to store the original biometric images. Reviewers
also suggested that we include ANSI*s Biometric Information Management and
Security (ANSI X9.84- 2001) standard in the report and the AAMVA driver*s
license standard that includes biometrics. Reviewers also stated we should
mention the ongoing biometric standards activities of the InterNational
Committee for Information Technology Standards Technical Committee M1,
Biometrics, which was established in November 2001. We incorporated these
suggestions into the report.
Some reviewers expressed concern that using biometrics as suggested in the
scenarios would be ineffective in preventing terrorists from entering the
United States. Some reviewers believed that the report needs to better
emphasize the limitations and operational concerns associated with
biometrics. For example, reviewers suggested that we highlight the fact
that adding biometrics to a travel document can bind only a person*s
claimed identity to the document. Additionally, the claimed identity is
only as good as the credentials that a person uses to claim that identity.
They also mentioned that biometric systems are not perfect and that
operational procedures must address weaknesses in any system
implementation.
We state that the goals of any biometric system for border security need
to be defined before any decision to design, develop, and implement it. We
describe a risk- based approach to security that can help with the
definition of goals. Part of this approach is an identification of
adversaries and threats. Regardless of whether terrorists are considered
the only adversary to border security, a vulnerability analysis is
required to
Chapter 6: Summary Page 135 GAO- 03- 174 Biometrics for Border Security
determine how an adversary can illegally enter the country. As we state in
the report, biometric technologies are not a panacea for border security
problems. Technology and people must work together to execute border
security processes. As reviewers have pointed out, increasing security at
the ports of entry does not address problems with people illegally
crossing into the United States at points other than official ports of
entry. Biometric technologies can only help support a single task, the
binding of an identity to an individual. Numerous other technologies and
people are needed to support other border security processes that,
together as a whole, protect the border. We have further adjusted the
report, where appropriate, to make this point clear.
Reviewers commented that our report does not take a sufficiently
forwardlooking approach to the civil liberties problems. The reviewers
believed that the report appears to downplay significant issues with the
effectiveness of biometric technologies and the significant civil
liberties issues surrounding the use of biometrics for border control.
These reviewers suggested that civil liberties issues need to be better
addressed, including the potential for unauthorized access to data, abuse
by those with authorized access, bad data in the system, the consequences
of false matches on individuals, the need for greater transparency, and
the dangers of racial or other profiling.
In the report, we summarize guidelines for addressing privacy in biometric
systems. Similar to the need to define the security goals of a biometric
border control system, there is a need to define the privacy requirements
for the system. The guidelines can help decision makers develop a policy
consensus on the amount of privacy to build into such a system. As we
point out, many of the issues surrounding the implementation of privacy
are not technical issues. Instead, they surround the management policies
governing the use of the technology and the information generated by such
a system.
We also received numerous technical comments on topics such as the
specific enrollment numbers for biometric applications, template sizes,
the maturity of new technologies, and equipment costs. We have
incorporated these comments, where appropriate, in the report.
Appendix I: Our Technology Assessment Methodology
Page 136 GAO- 03- 174 Biometrics for Border Security
The objectives of this technology assessment were to 1. Identify biometric
technologies currently deployed, currently available
but not yet deployed, or in development that could be deployed in the
foreseeable future for use in securing the nation*s borders.
2. Determine how effective these technologies are for helping provide
security to our borders currently or are likely to be in the future.
3. Determine the economic and effectiveness trade- offs of implementing
these technologies.
4. Identify the implications of biometric technologies for personal
security and the preservation of individual liberties.
To identify and describe biometric technologies, we convened, with the
assistance of the National Academy of Sciences (NAS), two meetings on
biometrics and border control issues, which included manufacturers of
facial, fingerprint, and iris recognition and hand geometry technologies.
1 The meetings also included informed representatives from academia,
government, and industry groups; privacy and civil liberty advocates; and
other stakeholders such as representatives of border communities and trade
organizations. We interviewed manufacturers of biometric technologies and
reviewed their publications to obtain descriptive information about their
equipment. We interviewed officials from biometric industry organizations,
including the Biometric Consortium and the Biometric Foundation. We also
interviewed the consulting firm the International Biometric Group (IBG).
We attended the biometrics session of the International Industrial
Security Conference, where technologies were demonstrated, and we
discussed various aspects of the technologies with industry
representatives.
To identify the current deployment of biometric technologies, we conducted
a literature search and reviewed reports of deployments, tests, and pilots
of biometric technologies. We interviewed certain users of biometric
technologies, including the Federal Bureau of Investigation (FBI),
Immigration and Naturalization Service (INS), National Security
1 We have a standing contract with NAS under which NAS provides assistance
in convening groups of experts to provide information and expertise to our
engagements. NAS uses its scientific networks to identify participants and
uses its facilities and processes to arrange the meetings. Recording and
using the information in a report is our responsibility. Appendix I: Our
Technology Assessment
Methodology
Appendix I: Our Technology Assessment Methodology
Page 137 GAO- 03- 174 Biometrics for Border Security
Agency (NSA), National Institute of Standards and Technology (NIST), the
Department of State, and the Canada Customs and Revenue Agency.
To determine how effective the technologies would be in helping to secure
the nation*s borders, we needed to understand the current border security
environment. We reviewed relevant statutes and regulations and interviewed
State Department and INS officials at headquarters and INS officials at
three ports of entry: El Paso, Texas (land); Miami, Florida (air and sea);
and Niagara Falls, New York (land). We reviewed and analyzed statistics
from INS*s Performance Analysis System for fiscal year 2001.
To determine the effectiveness of biometric technologies, we reviewed test
documentation from academic, government, and industry sources. In
particular, we interviewed and reviewed documentation from the Department
of Defense (DOD), Federal Aviation Administration (FAA), INS, NIST, Sandia
National Laboratories, the State Department, the United Kingdom*s National
Physical Laboratory (NPL), and IBG.
To determine the economic and effectiveness trade- offs of implementing
biometric technologies, we identified four different scenarios for
implementing them and built cost models to obtain life- cycle costs for
each scenario. The cost models represent rough order of magnitude costs
and are based on DOD*s cost element structure for major automated
information systems. To build the cost models, we used data provided by
the FBI, IBG, Naval Center for Cost Analysis, State Department, and
various vendors. We reviewed the cost model and assumptions associated
with each model with IBG and the State Department and incorporated their
feedback where appropriate.
In addition, we performed uncertainty analysis on the cost models, using a
Monte Carlo simulation tool called Crystal Ball to analyze the effects of
varying inputs and outputs of the modeled scenarios. This allowed us to
try multiple what- if scenarios with our spreadsheet cost model values and
cells. We used the results of this analysis to provide a probability value
for our point estimates, as well as to provide a risk- adjusted life-
cycle cost estimate for each scenario. Crystal Ball examines the degree of
risk in forecasts by using Monte Carlo simulation techniques that forecast
all statistically possible results for a given situation. We applied a
probability distribution to each parameter that we thought could vary,
such as the costs for development and installation, annual operating
personnel, and additional square feet in embassy or consular facilities.
Then, Crystal Ball generated random values for each cell, according to the
parameters we
Appendix I: Our Technology Assessment Methodology
Page 138 GAO- 03- 174 Biometrics for Border Security
chose to represent the risk. The software displayed the distribution of
results, showing the highest, lowest, and most likely values.
We analyzed the benefits of each scenario and described the effects on
people and processes, based on our understanding of the technology and
current border control processes.
To determine the implications of biometric technologies, we reviewed
relevant statutes and regulations and interviewed officials from privacy
and civil liberty groups. We also heard from representatives of these
groups at our meetings convened by NAS. We met with the Greater El Paso
Chamber of Commerce to obtain its thoughts on the introduction of
biometrics and the potential economic effect in the El Paso area. We
reviewed data from the Department of Commerce and Department of
Transportation to determine the potential economic effect of implementing
biometrics.
We provided a draft of this report to the Department of State and
Department of Justice for their review. We include their comments in
appendixes VII and VIII, respectively. In addition, we provided a draft of
this report to selected attendees of the two meetings NAS convened for
this work and other interested organizations.
Three recognized independent external reviewers reviewed our process for
conducting our work. In addition to providing a sound analysis of this
assessment, the reviewers made recommendations for improving and enhancing
future assessments should the Congress ask us to do more in the future.
We conducted our work from March to October 2002 in Washington, D. C.;
Clarksburg, West Virginia; El Paso, Texas; New York, New York; Niagara
Falls, New York; Miami, Florida; and Philadelphia, Pennsylvania. We
performed our work in accordance with generally accepted government
auditing standards.
Appendix II: Fingerprint Recognition Technology
Page 139 GAO- 03- 174 Biometrics for Border Security
Fingerprint identification has two basic premises. The basic
characteristics of fingerprints do not change with time* persistence* and
each person*s fingerprints are unique* individuality. Scientific studies
in the mid- 1800s established the persistence of friction ridge patterns
on human fingers, beginning in the embryonic stage and extending
throughout life, except for accidental damage.
Manual inspection of millions of fingerprints has led to the widely
accepted notion of fingerprint individuality. However, it has not been
formally established by scientific means that a person*s fingerprints are
unique. Because it is impossible to obtain the fingerprints of every
person in the world, estimating fingerprint individuality requires
statistical methods to project the probability that two people will have
the same fingerprint. The FBI*s Integrated Automated Fingerprint
Identification System (IAFIS) is the largest biometric database in the
world with its 400 million fingerprints. Although the FBI has never
discovered matching fingerprints from two individuals, tests have not been
performed to conclusively verify that the fingerprints in IAFIS are
unique.
In response to the need for a study to rigorously test the scientific
basis of fingerprint individuality, the National Institute of Justice
issued a formal solicitation in March 2000 for *Forensic Friction Ridge
(Fingerprint) Examination Validation Studies.* The objectives were basic
research to measure the amount of detail in a single fingerprint that can
be used for comparison and the amount of similar detail between two
separate fingerprints.
Fingerprint identification has been used in law enforcement over the past
hundred years and has become the de facto international standard for
positively identifying individuals. The FBI has been using fingerprint
identification since 1928. The first fingerprint recognition systems were
used in law enforcement about four decades ago. Advances in optical
scanning technology since the 1980s have made the technology practical for
noncriminal applications. Figures 26 through 28 illustrate some current
applications of fingerprint recognition technology. Appendix II:
Fingerprint Recognition
Technology
Appendix II: Fingerprint Recognition Technology
Page 140 GAO- 03- 174 Biometrics for Border Security
Figure 26: Using Fingerprint Biometrics for Physical Access
Source: National Coordination Office for Information Technology Research
and Development.
Figure 27: Using Fingerprint Biometrics for Logical Access
Source: Identix Incorporated.
Appendix II: Fingerprint Recognition Technology
Page 141 GAO- 03- 174 Biometrics for Border Security
Figure 28: A Fingerprint Biometric Device for Personal Identification
Source: Sagem Morpho Inc.
The use of fingerprints for forensic evidence was challenged recently. In
1999, the defense in U. S. v. Mitchell pointed to the Daubert Opinion,
established in a 1993 U. S. Supreme Court case, that prompted the
scientific community to address questions about the reliability and
validity of certain types of evidence, such as whether the evidence has
been adequately tested, what its error rate is, and whether there are
standards for what constitutes a fingerprint match. 1 The U. S. Court of
Appeals in U. S. v. Mitchell held that fingerprinting meets the necessary
criteria for admissibility as evidence. More recently, in January 2002, in
U. S. v. Llera Plaza, the judge refused to allow fingerprint experts to
express an opinion that a particular latent print matched or did not match
the rolled print of a particular person. 2 However, in March 2002, the
judge reversed himself and concluded that the court*s use of expert
fingerprint identification testimony, subject to careful trial court
oversight, could be allowed.
1 U. S. v. Byron C. Mitchell (Criminal Action No. 96- 407- 1, U. S.
District Court for the Eastern District of Pennsylvania 1999). 2 U. S. v.
Carlos Ivan Llera Plaza, Wilfredo Martinez Acosta, and Victor Rodriguez
(Criminal Action No. 98- 362- 10, 11, 12, U. S. District Court for the
Eastern District of Pennsylvania 2002).
Appendix II: Fingerprint Recognition Technology
Page 142 GAO- 03- 174 Biometrics for Border Security
Fingerprint recognition technology uses the impressions made by the unique
ridge formations or patterns found on the fingertips. 3 The technology
uses two main types of fingerprints: flat and rolled. A flat fingerprint
is obtained by pressing the finger flat against the scanner, capturing an
impression of the central area between the fingertip and the first
knuckle. A rolled fingerprint is obtained by rolling the finger from one
edge of the fingernail across to the other, capturing an impression of the
side ridges as well. A flat fingerprint is quicker to capture, but a
rolled fingerprint can provide up to 50 percent more surface area for
future comparisons.
Whether flat or rolled, the image of the fingerprint is commonly captured
by a scanner based on optical, silicon, or ultrasound technology. Optical
technology is the oldest and most widely used; it requires that the finger
be placed on a coated platen, typically made of hard plastic. In most
devices, a charged coupled device converts the image, with dark ridges and
light valleys, into a digital signal. The brightness is adjusted
automatically or manually to produce a usable image. Although most
companies use optical technology, the trend is toward silicon.
One type of silicon technology is based on capacitance, where the silicon
sensor typically acts as one plate of a capacitor and the finger is the
other. The capacitance between the platen and the finger is converted into
an 8 bit gray- scale digital image. Although ultrasound technology is
potentially more accurate than either optical or silicon, its performance
has not been assessed in widespread use. It captures the fingerprint by
transmitting acoustic waves and measuring the distance by the impedance of
the finger, the platen, and air.
After a fingerprint image has been captured, it is enhanced to reduce
image noise, formed when a fingerprint is converted into a digital image;
the noise distorts the image, generally as repetitive patterns or random
dots. A fingerprint image is one of the noisiest of image types,
predominantly because fingertips become dirty, cut, scarred, creased, dry,
wet, and worn. Image enhancement reduces this noise and enhances the
definition of ridges and valleys. To allow for precise locations of ridge
features, ridges are thinned from an original width of 5 to 8 pixels down
to 1 pixel.
3 Ridges are the upper skin layer segments of the finger; valleys are the
lower segments. How the Technology
Works
Appendix II: Fingerprint Recognition Technology
Page 143 GAO- 03- 174 Biometrics for Border Security
For a gray- scale image, areas lighter than a set threshold may be
discarded, while darker areas may be made black. Image enhancement is
relatively time consuming, since a 500 x 500 pixel fingerprint image has
250,000 pixels and each pixel is enhanced. Consequently, many fingerprint
systems are designed to limit enhancement operations at this stage in
order to reach a match result quickly, trading faster match time for
poorer image quality.
Following image enhancement, several steps are required to convert a
fingerprint*s unique features into a template. Known as feature
extraction, this is the basis of fingerprint recognition technology and
the various vendors* proprietary algorithms. We discuss the different
algorithms below. In none of these methods is the template a full
fingerprint image, and a real fingerprint cannot be recovered from the
digitized template. The generated template ranges from 250 bytes for
minutiae- based templates to about 1,000 bytes for ridge- pattern- based
templates.
Approximately 80 percent of fingerprint recognition vendors base their
algorithms on minutiae points, or the breaks in fingertip ridges. A
typical fingerprint image may produce between 15 and 50 minutiae,
depending on the portion of the image captured. As shown in figure 29, the
most basic minutiae are ridge endings (where a ridge ends) and
bifurcations (where a single ridge divides into two). Minutiae- Based
Templates
Appendix II: Fingerprint Recognition Technology
Page 144 GAO- 03- 174 Biometrics for Border Security
Figure 29: Common Fingerprint Features
Source: GAO adaptation of FBI data.
Before minutiae can be identified, an algorithm must search the processed
fingerprint image and filter out distortions and false minutiae. False
minutiae can be caused by scars, sweat, or dirt and often create anomalies
that can typically be detected. For instance, minutiae that seem out of
place could include two ridge endings on a very short isolated line; the
line would probably stem from image noise. Similarly, endings at the
boundary of the fingerprint would be eliminated because they are not true
endings but, rather, the edge of the image captured by the scanning
device. A large percentage of minutiae candidates are discarded this way.
Once the minutiae are identified, their locations are usually set on an x,
y axis and their angles are measured (typically by the direction of a
ridge or valley ending). For each established minutiae point, neighboring
minutiae and the number of ridges in between are recorded. The result of
this stage is a minutiae template of the fingerprint. Because of
differences in the
Valley Ridge Ending Bifurcation Ridge
Appendix II: Fingerprint Recognition Technology
Page 145 GAO- 03- 174 Biometrics for Border Security
determination, placement, and analysis of minutiae points, no two
algorithms can be expected to yield the same template from a given
fingerprint.
In a verification system, templates are usually matched by comparing the
neighborhoods of nearby minutiae for similarity. If a comparison indicates
only small differences, the neighborhoods are said to match. Comparisons
are performed exhaustively for all combinations of neighborhoods, and if
enough similarities are found, the fingerprints are said to match. One
result of this verification stage is a match score, usually a number
between 0 and 1 (or 10 or 100). Higher values in the range indicate higher
confidence in a match, and the match score is measured against a
predetermined threshold. If the score is greater than the threshold, the
match result is said to be true. The threshold can be lowered to reduce
the number of false nonmatches, but the trade- off is a greater number of
false matches. Some systems score the difference between two templates, in
which case a lower score is considered a match.
In an identification system, which compares a trial fingerprint template
to an entire database, the verification technique described above would be
impractical. Making comparisons to every fingerprint in the database by
neighborhoods would lengthen computation time extensively. Instead, a two-
step process is typically used for 1: N matching. First, to provide an
indexing method, the trial fingerprint and the reference template in the
database are categorized according to an established fingerprint type
(such as the plain arch, loop, or plain whorl shown in figure 30). This
step is called binning, in which a pattern comparison quickly eliminates
the bulk of the nonmatches. Care must be taken in binning. Errors in
assigning images to bin categories increase the likelihood of a false
nonmatch.
Figure 30: Established Fingerprint Types
Plain Arch Loop Plain Whorl Source: FBI.
Appendix II: Fingerprint Recognition Technology
Page 146 GAO- 03- 174 Biometrics for Border Security
In the second step for 1: N matching in identification, the trial template
is compared by minutiae neighborhood to each reference template that
closely matches the trial template pattern.
In matching ridge patterns, data are extrapolated from a particular series
of ridges, to be used in enrollment for the basis of future comparisons.
The ridge series are chosen so as to maximize the amount of unique
information that is recorded* for example, those with an unusual ridge
combination. At verification, a segment of the same area must be found and
compared. The match result contains information on how well the stored
images fit the verification image. This information is measured against a
threshold to determine whether the match result is true.
Fingerprint recognition technology companies number more than 75. There
are more fingerprint recognition vendors than for all other biometrics
combined. Some of the leading companies are listed in table 20. Ridge-
Pattern- Based
Templ at es The Leading Vendors
Appendix II: Fingerprint Recognition Technology
Page 147 GAO- 03- 174 Biometrics for Border Security
Table 20: Leading Vendors of Fingerprint Recognition Biometrics Scanner
for image capture Vendor Optical Silicon Ultrasound Other a
ActivCard X Astro Datensysteme AG X AuthenTec Inc. X Bergdata Biometrics
GmbH X BIO- key International X BioLink Technologies International Inc. X
Biometric Access Corporation X Bioscrypt Inc. X X Cogent Systems Inc. X
Cross Match Technologies Inc. X Delsy X X DigitalPersona Inc. X Ethentica
X Fingerprint Cards AB X Identix Inc. X Infineon Technologies AG X
Polaroid Corp. X Precise Biometrics X SAGEM Morpho Inc. X SecuGen Corp. X
Siemens AG X Sony Corp. X X STMicroelectronics X Thales X Ultra- Scan
Corp. X Veridicom Inc. X a Includes middleware and emerging scan
technologies that use polymer or fiber optic readers.
Source: GAO analysis of vendor data.
Fingerprint readers designed for physical access control range from about
$1,000 to $3,000 per unit. Software licenses for the fingerprint
technology are about $4 per user enrolled. For smaller fingerprint
scanners, maintenance costs are between 15 percent and 18 percent of cost.
A live scan 10- print fingerprint reader costs about $25,000. The
maintenance cost of the larger machines is approximately 14 percent of the
cost of the reader.
Although fingerprints are stable physiological characteristics, daily wear
can cause the performance of some fingerprint recognition technologies to
The Cost of Devices
Performance Issues
Appendix II: Fingerprint Recognition Technology
Page 148 GAO- 03- 174 Biometrics for Border Security
drop drastically. Although high- quality enrollment improves long- term
performance, the fingerprints of about 2 to 5 percent of people cannot be
captured because the fingerprints are dirty or have become dry or worn
from age, extensive manual labor, or exposure to corrosive chemicals.
Also, IBG*s comparative biometric testing has shown that certain ethnic
and demographic groups (elderly populations, manual laborers, and some
Asian populations) have fingerprints that are more difficult to capture
than others*.
Optical and silicon scanning technologies have unique performance issues.
Scanning fingerprints optically can be prone to error if the platen has a
buildup of dirt, grime, or oil* producing leftover fingerprints from
previous users, known as latent prints. Severe latent prints can cause the
superimposition of two sets of prints and image degradation. Although
silicon scanners generally produce a higher- quality image, high- quality
fingerprint capture is more difficult because the sensor size is smaller
than that used in optical scanners. Ultrasound scanning technology is
designed to penetrate the dirt and residue on platens.
Optical and silicon scanners using minutiae- based and pattern- matching
technologies have been tricked into accepting reactivated latent prints or
artificial fingers with forged fingerprints. Latent fingerprints were
reactivated by simply breathing on the sensor or by placing a water-
filled plastic bag on the sensor*s surface. Latent fingerprints could also
be reconstructed and authenticated by dusting the sensor*s platen with
commercially available graphite powder and lifting with adhesive tape.
Artificial fingers made with candle wax or gelatin and the fingerprints of
enrolled individuals have also successfully fooled the system.
Because law enforcement agencies identify criminals with fingerprints, the
recognition technology*s similarity to forensic fingerprinting causes some
percentage of users discomfort. Privacy advocates fear that fingerprint
recognition systems may collect data for one purpose but then use the data
for other purposes, such as in forensic applications or for tracking
people*s activities. Also, people may have hygiene issues with touching
the plate of a scanner that many people have touched. User Acceptance
Appendix II: Fingerprint Recognition Technology
Page 149 GAO- 03- 174 Biometrics for Border Security
The FBI*s IAFIS is an automated 10- fingerprint matching system that
relies on rolled fingerprints. The more than 40 million records in its
criminal master file are connected electronically with all 50 states and
some federal agencies. IAFIS was designed to handle a large volume of
fingerprint checks against a large database of fingerprints. It processes,
on average, approximately 48,000 fingerprints per day and has processed as
many as 82,000 in a single day. IAFIS*s target response time for criminal
fingerprints submitted electronically is 2 hours; for civilian fingerprint
background checks, 24 hours. According to FBI data from August 2002, the
majority of criminal fingerprints were answered in less than 7 minutes and
the majority of civilian fingerprints were answered in less than 4
minutes; 88.2 percent of criminal prints and 66.1 percent of civilian
prints were completed in less than 2 hours. For fingerprint submissions in
papercard format, the response time is 3 days between receipt and mailed-
back response. The FBI claims that IAFIS has a false match rate (FMR) of
about 1.5 x 10 -12 with a false nonmatch rate (FNMR) of about 1.5 to 2.0
percent. The failure to enroll rate (FTER) is about 0.5 percent for
criminal searches and about 2.5 percent for civilian background searches.
INS began developing the Automated Biometric Fingerprint Identification
System (IDENT) around 1990 to identify illegal aliens who are repeatedly
apprehended trying to enter the United States illegally. INS*s goal was to
enroll virtually all apprehended aliens. IDENT can also identify aliens
who have outstanding warrants or who have been deported. When such aliens
are apprehended, a photograph and two index fingerprints are captured
electronically and queried against three databases (see figure 31). One
database stores the fingerprints and photographs of approximately 300,000
aliens INS has previously apprehended; it tracks the number of
apprehensions. The second database stores the fingerprints and photographs
of approximately 240,000 criminal aliens convicted of an aggravated
felony, among other criteria. The third database stores the fingerprints
and photographs of more than 4 million illegal aliens who were
apprehended, enrolled in IDENT, and then permitted to voluntarily depart
the United States or to withdraw their applications for admission at ports
of entry. The fingerprint query of the three databases normally takes 2
minutes. In March 2002, the FMRs for the four fingerprint search types
were 4.05 percent for flat to flat, 2.60 percent for flat to roll, 0.70
percent for roll to roll, and 1.19 percent for roll to flat. The
Technology*s
Maturity Operational Uses
Appendix II: Fingerprint Recognition Technology
Page 150 GAO- 03- 174 Biometrics for Border Security
Figure 31: An IDENT Workstation
Source: INS.
A number of states (including Arizona, California, Connecticut, Illinois,
Massachusetts, New Jersey, New York, and Texas) require applicants for
welfare benefits to submit their fingerprints in order to eliminate
duplicate participation and to deter fraud. The first social service
fingerprint recognition system in the nation was the Los Angeles County
Automated Fingerprint Image Reporting and Match system, which enrolled
311,000 clients between 1992 and 1995. On the basis of a study group of
24, 000 fingerprint recipients, it was determined that about 7 percent of
the cases on the benefit rolls in Los Angeles were multiple identities. As
of January 2001, this was the only substantial finding of multiple
identity fraud in any of the various state welfare fingerprint programs.
Since January 31, 2002, immigrants seeking asylum in the United Kingdom
are issued an application registration card to allow for quick and
positive identification of all asylum applicants. The smart cards,
manufactured by SAGEM Morpho Inc., store two fingerprint templates on a
memory chip. An extension of the Immigration and Asylum Fingerprint System
(IAFS), they are intended to prevent fraudulent behavior, such as
impersonations to avoid removal or to make false benefits claims. The
United Kingdom plans to adopt the Eurodac Convention and Protocol, which
assists European Union members in applying the provisions of the Dublin
Convention, a framework for ensuring that an asylum claim is heard within
the European Union only once. Once the United Kingdom has
Appendix II: Fingerprint Recognition Technology
Page 151 GAO- 03- 174 Biometrics for Border Security
adopted the Eurodac Convention and Protocol, British IAFS fingerprint data
will be transmitted electronically to a central database, accessible to
other members for fingerprint comparisons. The data will be retained for
10 years, with the exception of fingerprints from asylum seekers who are
granted citizenship by European Union members. The prints of these new
citizens will be immediately erased.
FAA Fingerprint Recognition Testing
In 2001, FAA conducted operational testing of a fingerprint recognition
reader for access control by different groups of people in various
operating environments. Following the test, the biometric system was
removed. Enrolling users in the fingerprint reader system took an average
of 3 minutes and 30 seconds. Two of 38 users were unable to enroll because
of the poor quality of their fingerprints. Passing through the door took
about 10 seconds using fingerprint recognition, compared to an average of
about 2 seconds before the device was installed. Performance rates varied
at the three different security- level thresholds tested. The FNMR ranged
from about 6 percent to about 17 percent for closely controlled test
subjects. For actual airport employees using the door in a less controlled
environment, the FNMR ranged from about 18 percent to about 36 percent.
The FMR ranged from 0 percent at the highest security level to
approximately 8 percent at the lowest security level.
O*Hare International Airport, Chicago
In 1998, FAA funded an operational test at Chicago*s O*Hare International
Airport involving smart cards and fingerprint recognition identification
devices to screen employees of motor carrier and air cargo companies at
access control points to cargo areas. Truck drivers were instructed to
insert a smart card into the smart card reader and to confirm their
identity by placing the enrolled fingers on the fingerprint reader. 4
Fingerprints were chosen over other biometrics because of the users*
operational requirements, the perception that fingerprint recognition was
one of the least intrusive technologies, and the results from a 1997 study
that determined that fingerprints could be used in a variety of
applications in the trucking industry.
4 To allow for scarred or injured fingers, drivers typically enrolled two
digits. Pilots
Appendix II: Fingerprint Recognition Technology
Page 152 GAO- 03- 174 Biometrics for Border Security
Because all users* verification attempts were voluntary, only some of the
users who were initially rejected by the system chose to try again. For 65
users, the first- try FNMR was 28 percent. Of the seven users who chose to
try again, 71 percent successfully accessed the system. If all rejected
users had retried with this rate of success, only 8 percent of the users
would have been rejected after two tries. 5 Testing of impostors*
fingerprints against the operational database was not performed, so an FMR
could not be obtained for the O*Hare databases.
Fingerprint Verification Competition 2000
The Fingerprint Verification Competition 2000 (FVC 2000) tested relative
technology performance in a one- to- many application and was not intended
to predict performance of fingerprint recognition technology in a real
environment. Eleven algorithms were submitted, seven from academic groups
and four from companies (one from Ditto Information & Technology Inc., one
from FingerPin AG, and two from SAGEM Morpho Inc.). Three databases in
this competition were acquired in a laboratory environment, using a
variety of sensors (both optical and silicon), while the fourth contained
synthetically generated images. Enrollment time averaged 0.20 to 10.42
seconds, 10 of the algorithms requiring no more than 3.18 seconds.
Matching time averaged 0.20 to 2. 67 seconds, 9 of the algorithms
requiring no more than 1.58 seconds. The most accurate algorithm had an
average equal error rate (EER) of 1.73 percent, while the least accurate
algorithm had an average EER of 47.84 percent. These data are depicted in
table 21.
5 FNMR analysis from system performance testing by Jim L. Wayman, U. S.
National Biometric Test Center, College of Engineering, San Jose State
University, San Jose, California. Tests
Appendix II: Fingerprint Recognition Technology
Page 153 GAO- 03- 174 Biometrics for Border Security
Table 21: Summary of Results from the Fingerprint Verification Competition
2000 Participant Type Average EER Average
enroll time Average match time
SAGEM SA, France (Algorithm 1) Company 1.73% 3.18 1.22 SAGEM SA, France
(Algorithm 2) Company 2.28 1.11 1.11 Centre for Signal Processing, Nanyang
Technological University, Singapore Academic 5.19 0.20 0.20 CEFET- PR/
Antheus Technologia LTDA., Brazil Academic 6.32 0.95 1.06 Centre for
Wavelets, Approximation, and Information Processing, Department of
Mathematics, National University of Singapore, Singapore Academic 7.08
0.27 0.35 Kent Ridge Digital Labs, Singapore Academic 10.94 1.08 1.58
University of Twente, Electrical Engineering, Netherlands Academic 15.24
10.42 2.67 FingerPin AG, Switzerland Company 15.94 1.22 1.27 Inha
University, Korea Academic 19.33 0.71 0.76 Ditto Information & Technology
Inc., Korea Company 20.97 1.24 1.32 Natural Sciences and Mathematics,
Institute of Informatics, Macedonia Academic 47.84 1.44 1.71
Note: Time is in seconds. Source: FVC 2000.
Fingerprint Verification Competition 2002
Researchers from University of Bologna, Italy; San Jose State University,
California; and Michigan State University, East Lansing; jointly conducted
Fingerprint Verification Competition 2002 (FVC 2002), a large- scale
evaluation of fingerprint recognition technology that was a follow- up to
FVC 2000. There were 31 participants* 21 from companies, 6 from academic
institutions, and 4 others. As in FVC 2000, three databases were acquired
in a laboratory environment, using both optical and silicon sensors, and a
fourth contained synthetically generated images. Enrollment time averages
ranged from 0.11 to 7.05 seconds, 24 of the
participants requiring no more than 1 second. Matching time averages
ranged from 0.18 to 5.01 seconds, 21 of the participants requiring no more
than 1 second. The most accurate algorithm had an average EER of 0.19
percent, while the least accurate algorithm had an average EER of 50
percent. Table 22 depicts these data.
Appendix II: Fingerprint Recognition Technology
Page 154 GAO- 03- 174 Biometrics for Border Security
Table 22: Summary of Results from the Fingerprint Verification Competition
2002 Participant Type Average EER Average
enroll time Average match time
Bioscrypt Inc., United States (Algorithm 1) Industrial 0.19% 0.11 1.97
Anonymous Industrial 0.33 2.12 1.98 Anonymous Industrial 0.41 1.23 1.13
Bioscrypt Inc., United States (Algorithm 2) Industrial 0.77 0.07 0.22
Siemens AG, Germany Industrial 0.92 0.48 0.52 Neurotechnologija Ltd.,
Lithuania Industrial 0.99 0.56 0.56 SAGEM, France (Algorithm 1) Industrial
1.18 4.05 1.65 Andrey Nikiforov (Independent Developer), United States
Other 1.31 0.81 1.23 SAGEM, France (Algorithm 2) Industrial 1.42 0.77 0.66
Deng Guoqiang (Independent Developer), China Other 2.18 0.17 0.48 IDENCOM
AG, Switzerland Industrial 2.22 0.52 0.62 Suprema Inc., Korea Industrial
2.50 0.54 0.63 Anonymous Industrial 3.31 0.53 0.65 Biometrics System Lab,
Beijing University of Posts and Telecommunications, China Academic 3.76
0.57 0.59 Anonymous Industrial 4.19 0.18 0.18 HZMS Biometrics Co. Ltd.,
China Other 4.24 0.65 0.66 ActivCard Canada, Canada Industrial 5.21 0.68
1.76 Antheus Tecnologia Ltda, Brazil Industrial 5.46 0.20 0.54 TeKey
Research Group, Israel Industrial 5.72 0.01 3.15 FINGERPIN AG, Switzerland
Industrial 6.05 0.48 0.77 Inha University, Korea Academic 6.07 0.80 0.84
Aldebaran Systems, United States Industrial 6.16 1.81 1.81 Digital
Fingerpass Corporation, China Industrial 6.40 0.49 0.50 DATAMICRO Co.
Ltd., Russia Industrial 6.72 0.33 0.56 Anonymous Industrial 7.12 0.24 0.28
Department of Computer Science and Information Engineering, Da- Yeh
University, Taiwan Academic 9.04 0.13 0.15 Anonymous Industrial 12.09 0.68
0.70 AILab, Institute of Automation, The Chinese Academy of Sciences,
China Academic 14.66 0.57 0.65 University of Tehran, Electrical and
Computer Department, Iran Academic 16.79 1.16 1.19 Anonymous Other 39.10
0.52 0.63 Anonymous Academic 50.00 7.05 5.01
Note: Time is in seconds. Source: FVC 2002.
Appendix II: Fingerprint Recognition Technology
Page 155 GAO- 03- 174 Biometrics for Border Security Biometric Product
Testing
NPL conducted a performance evaluation of seven different biometric
systems from May to December 2000. The fingerprint part of the test
included two types of systems, one based on optical technology and the
other on silicon technology. The vendor of the silicon sensor was
VeriTouch Ltd., with alternative enrollment and matching algorithms
provided by Infineon Technologies AG. The vendor of the optical sensor was
not identified.
The silicon system*s FTER was 1.0 percent, the optical system*s 2. 0
percent. FMR and FNMR measure the accuracy of the matching process.
Adjusting the decision criteria can make for a trade- off between false
match and false nonmatch errors. At an FMR of about 2 percent, the FNMR
was about 4.3 percent for the silicon sensor with the alternative
algorithm. Additional experimental results are summarized below:
The silicon sensor system had a mean transaction time of 19 seconds, a
median of 15 seconds, and a minimum of 9 seconds. The optical fingerprint
system had a mean transaction time of 9 seconds, a median of 8 seconds,
and a minimum of 2 seconds.
The silicon sensor system could make 60 matches per minute, the
alternative algorithm 2,500 matches per minute. The optical system could
make only 50 matches per minute. These diagnostic programs had significant
overheads, so the matching algorithm may be significantly faster than the
results showed, perhaps by a factor exceeding 100.
For both the silicon sensor and the optical system, younger people
generally had a lower FNMR than older people, and the FNMR for attempts
made immediately following enrollment were lower than those made on second
or third visits.
Republic of the Philippines Social Security System Identification Card
Benchmark Test
In May 1997, the U. S. National Biometric Test Center at San Jose State
University conducted an automated fingerprint identification system (AFIS)
benchmark test for the Republic of the Philippines Social Security System
Identification Card Project. The test measured single comparison FMRs and
FNMRs (among other metrics) for each of the four participating
international AFIS vendors. The FMR and FNMR for each vendor were
Appendix II: Fingerprint Recognition Technology
Page 156 GAO- 03- 174 Biometrics for Border Security
determined by matching 4,128 test images against a database of 4,080
fingerprints. Flat prints of the thumb through the ring finger of each
hand were collected from adult employee volunteers in the Social Security
System. One vendor returned 16 million cross comparisons with only one
false match, indicating a 95 percent statistical confidence in an FMR of
fewer than 3 in 10 million prints but with an FNMR approaching 20 percent.
Statistical analysis of the test results supports the feasibility of an
AFIS system that could support 16 million flat fingerprint comparisons
without a false match.
INS*s IDENT Benchmark Test
In June 1998, an independent verification and validation test was
conducted on the Cogent PMA3 Matcher configuration that was later
installed for IDENT. This benchmark test used a fingerprint test database
created by INS and provided to Cogent Systems Inc. that consisted of
129,712 rolled fingerprints, 951,956 flat fingerprints, and six different
search fingerprint image input files. The data were highly representative
of IDENT criminal alien records at the time. As about half of the search
subjects in the input files had mates in the rolled or flat fingerprint
database, the benchmark data were designed to obtain results with a high
confidence level.
All four types of searches (flat- flat, roll- flat, flat- roll, and roll-
roll) in operation in IDENT were tested during the benchmark. Two
verification match tests (flat- flat and flat- roll) were also conducted.
The results are displayed in table 23.
Appendix II: Fingerprint Recognition Technology
Page 157 GAO- 03- 174 Biometrics for Border Security
Table 23: INS*s IDENT Fingerprint Benchmark Test Results, 1998
Identification Verification Match test FNMR FMR FNMR FMR
Flat to flat 0.4% 5.4% 0 0. 1% Roll to flat 8.4 1. 5 a a Flat to roll 7.3
0. 2 1.6% 0.1 Roll to roll 0.2 0. 1 a a a Data not available.
Source: GAO analysis of INS data.
The results of the flat- to- flat and roll- to- roll search test were more
accurate than those of the mixed media tests* roll- to- flat and flat- to-
roll* because the mixed media searches resulted in a higher FNMR. The
results of the verification match test supported the use of the algorithm
for future INS verification applications.
The CANPASS* Airport pilot at Vancouver International Airport was
initiated in October 1995 using both fingerprint recognition and hand
geometry technologies. The pilot used identity cards and biometric
identification devices to allow previously screened travelers to bypass
customs and immigration lines. Qualified Canadian and U. S. residents
entered Canada through a special line by opening an automated gate with an
encoded identification card and providing a fingerprint or hand geometry
biometric for one- to- one authentication. Roughly a thousand travelers
registered with CANPASS in the pilot*s first 7 months, with an average
enrollment time of 15 minutes. Of 1,385 authentication attempts, 87
percent were successful and 13 percent were falsely rejected by the
technology and had to be processed manually. On the basis of these
results, authorities decided to use solely hand geometry for CANPASS*
Airport. Border Control
Applications Piloted and Deployed
CANPASS* Airport
Appendix II: Fingerprint Recognition Technology
Page 158 GAO- 03- 174 Biometrics for Border Security
The biometric border crossing card project is a joint effort of the
Department of State and INS to replace the paper- based card previously
issued to Mexican citizens. The new card is a laser visa, a credit- card-
like document, that permits the holder to enter the United States without
being issued further documentation for business or pleasure and to stay
for 72 hours or less, going no farther than 25 miles from the border. With
additional documentation, the laser visa can permit longer stays and
travel farther than 25 miles from the border. The laser visas are
manufactured by LaserCard Systems Corp. and are made of polycarbonate
material with a rectangular strip of optical memory for data storage. The
cards store a frontal facial image and the templates of two index flat
fingerprints. More than 5 million cards have been issued, but the
performance of the fingerprint recognition technology has not been
measured, because ports of entry have no scanners for reading travelers*
fingerprints and matching them with the information on the laser visa. INS
is buying and installing 30 readers at six ports of entry for a pilot
test.
About 250,000 people cross the Hong Kong* Shenzhen border daily, causing
long lines at the immigration checkpoint. The Hong Kong government plans
to issue new identity smart cards to residents in 2003. The smart cards
will hold a template of a rolled fingerprint to be matched against the
bearer at a self- service kiosk. The $21 million smart card contract was
awarded in March 2002, and distribution to the 6.8 million Hong Kong
residents will be phased in over 4 years.
The size of an identification system*s projected database has a
significant effect on the system*s configuration and cost. The larger the
database, the more storage devices are required. In addition, it takes
longer to search a larger database unless matching processor power is also
increased. Database size can also affect a system*s accuracy. Some
matching algorithms are effective only with relatively small databases and
are simply not capable of accurate matching against the larger numbers of
records found in forensic automated fingerprint identification systems.
Capturing fingerprints has been a significant issue in border control
pilots. In an unattended environment, trained users have generally skewed
fingers or have not pressed hard enough on the platen. The difficulty of
acquiring a usable fingerprint after three attempts has resulted in
approximately a 50 percent rejection rate. In addition, fingerprint
readers do not operate below freezing temperature, so the issues of
freezing and condensation are significant in selecting biometric systems.
Border Biometric Program
and Border Crossing Card Hong Kong Resident Smart Cards
Processing Issues Device Durability and Environmental Constraints
Appendix III: Hand Geometry Technology Page 159 GAO- 03- 174 Biometrics
for Border Security
Patents for hand geometry technology were first issued in the late 1960s
and early 1970s. The technology is based on the premise that the hand*s
bone structure, while changing over time, remains characteristically the
same. The hand*s shape usually stabilizes at age 13 or 14.
Hand geometry technology uses the hand*s distinctive features,
particularly the height and width of its back and fingers, to verify a
person*s identity. In measuring size and shape, a hand geometry system
collects more than 90 dimensional measurements, including finger width,
height, and length; distances between joints; and knuckle shapes. Although
the shape and size of our hands are reasonably diverse, they are not
necessarily unique. In larger populations, for example, it is almost
certain that various people have very similar hand dimensions.
Consequently, the technology cannot be used for 1: N identification.
In the measurement of the different features, a person places his or her
hand flat on the device*s metal surface, where pegs guide the fingers into
position. Hand geometry systems require a person to squeeze his or her
fingers against the pegs to prevent spoofing. Cameras capture two
orthogonal two- dimensional images of the back and sides of the hand (see
figure 32).
Figure 32: Fingers Guided by Pegs in a Biometric Hand Geometry Measurement
Source: Michigan State University, Biometrics Research Group.
Appendix III: Hand Geometry Technology How the Technology Works
Appendix III: Hand Geometry Technology Page 160 GAO- 03- 174 Biometrics
for Border Security
Only the spatial geometry is acquired; prints of the palm and fingers are
not taken. The derived template is 9 bytes in size, the smallest in the
biometric industry. In a process known as template averaging, the template
is automatically updated whenever the difference between the individual*s
hand and his or her reference template exceeds a designated threshold.
Recognition Systems Inc. (RSI) dominates the market in hand geometry
technology. Its systems are used in nearly every current implementation.
Companies that integrate hand geometry technologies include Electronic
Data Systems Corp. and ADT. However, Dermalog, a German company, is
developing an alternative technology that uses a pegless device. Biomet
Partners, a Swiss company, sells a finger geometry device that operates on
the same basic principles as the RSI hand geometry devices.
Hand geometry devices generally cost between $2,000 and $4,000. Training
is minimal, and no personnel costs are incurred because most hand geometry
devices are typically unattended.
Hand geometry disregards fingernails and surface details such as
fingerprints, lines, scars, and dirt. Except for jewelry, arthritis, water
retention, and swelling from pregnancy or hand injury, the hand is not
susceptible to major changes that would affect the technology*s accuracy.
However, because measurements of the hand are not distinct over a large
population, false matches can occur. Therefore, hand geometry is not
effective in large- scale1: N applications or in applications where
resistance to impostors is essential.
Hand geometry is generally perceived as nonintrusive, nonthreatening, and
noninvasive, and it bears very little of the stigma of other biometric
technologies. It lacks the forensic association that may affect users*
perceptions of fingerprint recognition systems. It is considered easy to
use, although a minimal amount of training may be required to learn how to
align the hands in the device. However, some people are uncomfortable
touching a device that many people have previously touched. The Leading
Vendors
The Cost of Devices Performance Issues
User Acceptance
Appendix III: Hand Geometry Technology Page 161 GAO- 03- 174 Biometrics
for Border Security
Hand geometry is an established, mature, and reliable technology that has
remained unchanged for several years. Hand geometry systems have been
deployed since the 1980s in tens of thousands of locations for access and
entry control, personal identification, and time and attendance
applications. For example, hand geometry is the most commonly deployed
biometric technology for controlling physical access and for processing
time and attendance records. Devices used for time and attendance
applications are often tied into physical access control systems. Hand
geometry devices have been installed at the entrances to more than half
the nuclear power plants in the United States. In 1991, San Francisco
International Airport installed hand geometry devices to protect secure
areas such as the tarmac and loading gates. At the 1996 Olympic Games in
Atlanta, Georgia, athletes used a hand geometry system to gain access to
Olympic Village.
FAA Hand Geometry Testing
In 2001, FAA and the National Safe Skies Alliance evaluated the
effectiveness of hand geometry technology for the use of access control of
airport employees. Following the test, the biometric system was removed.
Of the 39 people who successfully enrolled, 27 enrolled in an average of
57 seconds. The hand geometry system had varying security- level settings,
resulting in differing performance rates at verification. The FNMR ranged
from approximately 5 percent at a high security- level setting to less
than 1 percent at a low security- level setting. The FMR ranged from
approximately 0 percent at the high security- level setting to about 2
percent at the low security- level setting. Before the biometric
technology was installed, passing through the door was estimated at 2
seconds; after installation, the time increased to 8 seconds.
The results of testing under abnormal conditions are summarized below:
At the default security level setting, adding or removing rings similar
to the wide- band ring used in this test would very likely cause users to
be rejected at a high rate. Smaller rings do not appear to cause a higher
FNMR. The Technology*s
Maturity Operational Uses
Test s
Appendix III: Hand Geometry Technology Page 162 GAO- 03- 174 Biometrics
for Border Security
Wearing gauze pads or splints to cover injuries would also probably
cause a higher rejection rate. Standard adhesive bandages three- quarter
inches wide do not appear to cause higher FNMRs.
High backlight conditions did not noticeably affect FNMR.
Biometric Product Testing
From May to December 2000, NPL evaluated seven different biometric
technologies in a real- world environment for positive verification
comparative testing. The hand geometry portion of the test used RSI*s Hand
Key II, which had the fastest transaction time of the biometric
technologies compared. With 200 people enrolled, the FTER was 0 percent.
At an FMR of about 1 percent, the hand geometry system had an FNMR of
approximately 1. 4 percent. Additional experimental results were
The Hand Key II had a mean transaction time of 10 seconds, a median of 8
seconds, and a minimum of 4 seconds.
The matching algorithm could make 80,000 matches per minute when using a
SunUltra5 with a SunOS5.8 operating system, 270 MHz processor, and 128 MB
of memory.
Males had a somewhat lower FNMR than females.
Sandia National Laboratories
In 1991, Sandia National Laboratories evaluated five biometric
technologies, with nearly a hundred volunteers testing each technology.
Nearly 20,000 transactions were recorded for RSI*s ID- 3D hand geometry
devices. Overall, the hand geometry technology was the fastest, most
accurate, and most user- friendly device. Average verification time was 5
seconds, and the EER was about 0.2 percent. At the test threshold value,
the three- try FNMR was less than 0.1 percent, and the one- try FMR was
0.1 percent.
Sandia National Laboratories performed a field analysis with hand geometry
for physical building access control from 1993 to 1995. RSI*s model ID- 3D
HandKey biometric verifier was tested. Overall, 316 people used the device
in more than 100,000 instances. Sandia concluded that the device operated
differently in an exterior, unattended field installation than in previous
laboratory experiments: 7.20 percent of the individuals failed in the
first verification attempt, 53.48 percent in the second, and
Appendix III: Hand Geometry Technology Page 163 GAO- 03- 174 Biometrics
for Border Security
66.49 percent in the third. These percentages are not equivalent to FNMRs,
since not enough information was available to determine whether users who
were rejected should have been accepted. They may have been correctly
rejected because they may not have been who they claimed to be.
Researchers also found that maintenance and cleaning were paramount; when
the readers were not cleaned properly, performance was severely degraded.
Hand geometry is being used in many border control environments. The INS
Passenger Accelerated Service System (INSPASS), installed at seven U. S.
and two Canadian airports, uses 29 hand geometry kiosks to reduce
inspection time to less than 15 seconds for trusted travelers (see figure
33). INSPASS enrollment is open to all citizens of the United States,
Canada, Bermuda, and visa waiver countries. To enroll, travelers must
provide a passport or travel document and two fingerprints and present
their hand geometry biometric. Border Control
Applications Piloted and Deployed
INSPASS
Appendix III: Hand Geometry Technology Page 164 GAO- 03- 174 Biometrics
for Border Security
Figure 33: A Traveler Using an INSPASS Hand Geometry Device
Source: IR* Recognition Systems.
Since 1998, Ben Gurion Airport in Tel Aviv, Israel, has installed 21 hand
geometry kiosks and enrolled more than 100,000 passengers. The
implementation was initially offered only to frequent international
travelers, but passenger demand led to its expansion to all Israeli
citizens. Each month, more than 50,000 travelers use the automated
passenger screening system to reduce the immigration process to about 15
seconds. When using the system, a traveler swipes a magnetic stripe card
over a biometric reader (see figure 34). More than 2 million inspections
have been performed, and they are growing at 2 percent a month. In
addition to biometric authentication, the system checks the biometric
against Israeli law enforcement and immigration databases. Ben Gurion
Airport
Appendix III: Hand Geometry Technology Page 165 GAO- 03- 174 Biometrics
for Border Security
Figure 34: A Traveler Using Ben Gurion Airport*s Biometric Hand Geometry
System
Source: IR* Recognition Systems.
Headed by Electronic Data Systems, the Basel Project will implement a
system using facial recognition and hand geometry for day workers crossing
into and out of Israel from the Gaza Strip. Fingerprint technology was
rejected because the primary users are laborers whose fingerprints are not
reliable for biometric matching. People will enroll at the IsraelPalestine
land border, receiving a contactless smart card with a highresolution
picture and a hand geometry biometric. When entering or leaving Israel,
they will be processed through 42 routing passages to unattended
checkpoints at verification terminals inside a building. It is estimated
that 60,000 verifications will be processed daily in one- to- one matches
against stored templates in a central server, with a backup stored on the
smart card. Basel Project
Appendix III: Hand Geometry Technology Page 166 GAO- 03- 174 Biometrics
for Border Security
In June 1999, the Port of Rotterdam, Europe*s busiest container port,
implemented a hand geometry system designed to speed cargo movement. Each
truck driver*s identity is verified with the biometric template stored on
a radio frequency smart card, accessed through the truck*s window. It has
more than 6,000 users and has logged more than 3 million transactions.
The CANPASS* Airport pilot at Vancouver International Airport was
initiated in October 1995 using both fingerprint recognition and hand
geometry technologies. The pilot used identity cards and biometric
identification devices to allow previously screened travelers to bypass
customs and immigration lines. Qualified Canadian and U. S. residents
entered Canada through a special line by opening an automated gate with an
encoded identification card and providing a fingerprint or hand geometry
biometric for one- to- one authentication. The system*s use was
discontinued on September 11, 2001.
Hand geometry is well suited for most environments. The equipment is
durable and can withstand most workload demands. Various types of hand
geometry devices on the market are suitable for all types of climates (see
figures 35 and 36). Most can withstand temperatures ranging from *45
degrees to 120 degrees Fahrenheit and can provide protection against snow,
sleet, rain, splashing water, hose- directed water, falling dirt, and
wind- blown dust. Port of Rotterdam
CANPASS* Airport Device Durability and Environmental Constraints
Appendix III: Hand Geometry Technology Page 167 GAO- 03- 174 Biometrics
for Border Security
Figure 35: A Typical Hand Geometry Recognition Device
Source: IR* Recognition Systems.
Appendix III: Hand Geometry Technology Page 168 GAO- 03- 174 Biometrics
for Border Security
Figure 36: A Hand Geometry Recognition Device That Is Enclosed
Source: IR* Recognition Systems.
Appendix IV: Facial Recognition Technology Page 169 GAO- 03- 174
Biometrics for Border Security
Every day, people identify other people by their faces. Much research has
yielded evidence that people may recognize others* faces through a unique
process that highlights the importance of the location and shape of eyes,
nose, and eyebrows and face shape, chin, lips, and mouth, in decreasing
order. Because this process differs from how we recognize other objects,
the idea that machine recognition systems should also be face- specific
may have been encouraged. However, just as some people may have difficulty
differentiating between identical twins and other people with similar
features, facial recognition technology also cannot effectively
distinguish between people who resemble one another, and it still requires
development to full maturity. Nevertheless, active research over the past
10 years has made the technology commercially available.
Facial recognition identifies people by the sections of the face that are
less susceptible to alteration* the upper outlines of the eye sockets, the
areas around the cheekbones, the sides of the mouth. Systems using this
technology capture facial images from video cameras and generate templates
for comparing a live facial scan to a stored template. Facial recognition
technology can also be used to compare static images, such as digitized
passport photographs.
The comparisons are used in verifying and identifying individuals.
Verification systems compare a person*s facial scan to a stored template
for that person and can be used for access control. In an identification
system, a person*s facial scan is compared to a database of multiple
stored templates. This makes an identification system more suitable for
surveillance in conjunction with closed- circuit television (CCTV) to spot
suspected criminals whose facial characteristics have been captured and
stored in a database on a template. The face is the only biometric used in
a viable recognition technology that is able to operate without a user*s
cooperation, since a CCTV camera need only capture a picture for the
technology to generate a template. However, the technology is much more
able to identify people who are motivated to use the system correctly than
those who are uncooperative and can avoid recognition by, for example,
using disguises or taking other evasive measures.
The primary facial recognition technologies are used for one- to- one as
well as one- to- many matching. Whether used for verification or
identification, the stored image templates must be kept up to date, since
appearances naturally alter with age. Appendix IV: Facial Recognition
Technology
How the Technology Works
Appendix IV: Facial Recognition Technology Page 170 GAO- 03- 174
Biometrics for Border Security
However, IBG*s testing has found that the core technology is highly
susceptible to falsely nonmatching users in one- to- one verifications and
to failing to identify enrolled users in one- to- many identifications.
Two primary types of facial recognition technology are used to create
templates. 1 Requiring as many as 1,300 bytes, or as few as 84 bytes, they
are local feature analysis (LFA) and the eigenface method.
Patented by Visionics Corp.* now Identix Incorporated* LFA uses dozens of
images from regions of the face, resulting in feature- specific fields*
eyes, nose, mouth, cheeks. The fields* relative locations are incorporated
so that the face can be represented as a topographical grid made up of
blocks of features. The features represented by these blocks and their
positions are used to identify or verify the face (see figure 37).
Figure 37: Local Feature Analysis: A Topographical Grid of Facial Regions
Source: Identix Incorporated.
1 Other facial recognition technologies based on thermal patterns below
the skin are not yet commercially viable. Local Feature Analysis
Appendix IV: Facial Recognition Technology Page 171 GAO- 03- 174
Biometrics for Border Security
Just as Washington, D. C., can be identified by describing its landmarks*
locations and their relative positions (e. g., the National Mall has the
U. S. Capitol building to its east, the Lincoln Memorial to its west, and
the Washington Monument and Smithsonian museums at its center), a person*s
face can be identified by the features defined by LFA. Small shifts in a
feature may cause a related shift in an adjacent feature and the
technology can accommodate these changes in appearance or expression (such
as smiling or frowning). Since LFA does not provide a global
representation of the face, it is rendered ineffective when a person tilts
his or her head from a direct frontal pose to more than about 25 degrees
horizontally or more than about 15 degrees vertically.
Eigenface, meaning roughly *one*s own face,* is a technology patented at
the Massachusetts Institute of Technology. Unlike LFA, the eigenface
method always looks at the face as a whole. A collection of facial images
is used to generate a set of two- dimensional, gray- scale images
(eigenfaces) to produce the biometric template (see figure 38). The vast
majority of faces can be represented by locating distinctive features from
approximately 100 to 125 eigenfaces. When a live image of a person*s face
is introduced, the system represents the image as a combination of
templates. This combination is compared with a set of stored templates in
the system*s database, and the degree of variance determines whether or
not a face is recognized.
Figure 38: Two- Dimensional, Gray- Scale Images of an Eigenface Template
Source: Baback Moghaddam, MIT Media Laboratory.
The Eigenface Method
Appendix IV: Facial Recognition Technology Page 172 GAO- 03- 174
Biometrics for Border Security
Modifications of the algorithms used in LFA and the eigenface method can
lead to variances that incorporate
Neural network mapping: Comparisons of a live facial image with a stored
template are based on unique global features rather than individual
features. When a false match is made, the comparison algorithm modifies
the weight given to certain features, such as shadows.
Automatic face processing: Facial images are captured and analyzed from
the distances and distance ratios between features, such as between the
eyes.
The leading algorithms are licensed by Identix Inc. (which merged with
Visionics in June 2002) and Viisage Technology. Identix uses local feature
analysis; Viisage*s algorithm is based on the eigenface method.
A facial recognition server controlling access at a facility with up to
30,000 persons would cost about $15,000. Depending on the number of
entrances with installed facial recognition technology, the cost of the
software licenses would range from about $650 to $4,500. As the size of
the database and the number of attempted matches increased, so would a
system*s cost.
In addition to the server and software licenses, a live- scan facial
recognition surveillance system includes CCTV surveillance (see figure
39). A fully integrated CCTV system for physical access surveillance can
cost from $10,000 to $200,000, depending on the size of the entrance and
the degree of monitoring required. For additional CCTV equipment, cameras
can cost between $125 and $500. Cameras with advanced features can cost up
to $2,300. The Leading Vendors
The Cost of Devices
Appendix IV: Facial Recognition Technology Page 173 GAO- 03- 174
Biometrics for Border Security
Figure 39: CCTV Surveillance Equipment
Source: Pelco.
The effectiveness of facial recognition technology is influenced heavily
by environmental factors, especially lighting conditions. Variations in
camera performance and facial position, expression, and features
(hairstyle, eyeglasses, beards) further affect performance. Accurate image
alignment is necessary for the leading facial recognition algorithms,
which rely on identifying eye positions. As a result, current facial
recognition technology is most effective when used in consistent lighting
with cooperative subjects in a mug- shot- like position* where hats and
sunglasses are removed and everyone looks directly at the camera one at a
time.
Attempts to spoof live- scan facial recognition systems have been
successful. In one test, trial images were obtained by downloading
unprotected reference facial images to a computer and by taking digital
pictures of an enrolled person. These images were displayed on a notebook
computer monitor and were successfully matched, granting testers access to
the system. A video of an enrolled person moving his head slightly left
and right also fooled the system. Performance Issues
Appendix IV: Facial Recognition Technology Page 174 GAO- 03- 174
Biometrics for Border Security
When used in a verification system for access control, facial recognition
technology is typically considered by users to be less intrusive than
fingerprint readers, iris scanners, and other biometric technologies. It
can recognize people at a distance and does not require users to pause and
interact with the equipment. However, some users are concerned that when
used as a surveillance tool, facial recognition technology can facilitate
tracking them without their consent. To address such concerns, specific
policies for using facial recognition in a surveillance application have
been suggested, including the following.
As with any technology, public understanding of the operation and uses of
electronic surveillance might mitigate fears that the government may be
tracking people*s whereabouts. Signs indicating the use of facial
recognition in surveillance systems should be prominently displayed, and
the government entity using facial recognition for surveillance should
provide as much information as possible to the public about the
technology*s purposes and capabilities.
Concerns have been raised about the possibility that facial recognition
surveillance systems can identify law- abiding citizens, not only
terrorists or violent criminals. A *no match, no memory* policy dictates
that a person*s image is saved only if a match is made to a record in a
watch list database.
One issue that could arise is the government*s handling of the data it
collects. Even if a no match, no memory policy has been implemented, a
retention policy should be followed that indicates the time period after
which the data will be erased. Similarly, the data should be securely
stored and maintained.
Concern about how facial technology surveillance will be used is often
related to fear that the technology*s capabilities will be abused. Facial
recognition systems must be used only for the purpose they were designed
for, and some form of active oversight should be implemented. A
cooperative effort between government officials and citizen oversight
committees would provide accountability. User Acceptance
Transparency No Match, No Memory Data Retention Oversight
Appendix IV: Facial Recognition Technology Page 175 GAO- 03- 174
Biometrics for Border Security
The largest implementation of Identix*s facial recognition technology is
the Mexican Federal Electoral Institute*s program to eliminate duplicate
voter registrations. This system helps the Institute prevent citizens from
voting more than once under different aliases. Facial recognition is used
to compare people with matching names to determine whether the faces also
match. The system*s database, first used in Mexico*s July 2000
presidential elections, contains about 60 million images.
The largest deployment of facial recognition for surveillance began in
1998 in Newham Borough, London, England, when Identix*s facial recognition
technology was introduced to 12 town center cameras to record activity and
decrease street robbery in an unsafe neighborhood. With three hundred CCTV
cameras, this system captures faces and compares them against a police
database of about a hundred convicted street robbers known to have been
active in the previous 12 weeks. When a face does not match, the image is
deleted; when a match is found, an operator checks the result. In August
2001, 527,000 separate faces were detected and operators confirmed 90
matches against the database.
Public approval of Newham*s system was judged by comparing the results of
opinion polls over the course of the implementation. When Identix*s facial
recognition technology was first introduced, 50 percent of local citizens
approved of the system. After about 2 years of operation, the technology
was credited with a 34 percent reduction in street robbery, and the user
approval rating rose to 90 percent. As the system has not led directly to
any arrests, the effect of facial recognition technology appears to
function largely as a deterrent to street crime in the monitored area.
In the United States, Viisage*s facial recognition technology is deployed
in 17 states to identify people with credentials or identification
documents under more than one name. The majority of the states* databases
consist of image templates from driver*s license photographs. Illinois*s
driver*s license database consists of about 10 million images and has the
capacity for another 15 million images. The technology can perform a one-
to- many match against this database in less than 15 seconds, and about
15,000 images are captured daily. The Technology*s
Maturity Operational Uses
Appendix IV: Facial Recognition Technology Page 176 GAO- 03- 174
Biometrics for Border Security
Facial recognition surveillance systems have been deployed in casinos
worldwide, performing one- to- many matching against a database of casino
offenders. Although the notable facial recognition implementations are in
surveillance applications, facial recognition systems have been deployed
in selected environments as a one- to- one verification solution for
physical and logical access. Some casinos use facial recognition for
employee time and attendance processing, while applications for automated
teller machine fraud prevention and security have been implemented in
grocery stores and gas stations.
Identix has been involved in four pilots that use facial recognition for
surveillance at U. S. airports. The pilots had different operating
scenarios to determine the relationship between the correct match rate*
that is, the rate of actual matches* and the FNMR. Video cameras that were
not hidden from travelers were set up near the airport metal detectors.
The pilots were designed at some airports so that travelers were
specifically instructed to stop and look at the cameras; travelers at
other airports were not given such instructions.
From the four pilots, Identix concluded that lighting was the primary
performance factor. It learned also that the correct match rate, and
therefore the FMR, is quickly compromised as the threshold is adjusted to
minimize the FNMR. The data are shown in table 24.
Table 24: Identix Airport Facial Biometric Pilot Results Airport Status
False
match rate False nonmatch rate Notes
Boston Logan International, Mass. Completed Not reported ~10% Viisage
technology was also piloted. Dallas/ Fort Worth International, Texas
Completed 1. 2% 6* 15 Two cameras were used; when a match was made,
the person*s image was dispatched to a central control room for further
investigation. Fresno Yosemite International, Calif. Ongoing 1*5 5* 15 A
liquid crystal display instructed each traveler when
to pause in front of a fixed camera and when to resume walking.
Palm Beach International, Fla. Completed 0. 3 45 The objective was to
obtain an FMR as close to 0 as possible.
Source: GAO analysis of Identix data.
From May to June 2000, Naval Surface Warfare Center, Crane Division,
evaluated an identification system in the Facial Recognition Vendor Test
2000 (FRVT 2000). The two test categories conducted during the Pilots: U.
S. Airport
Surveillance Facial Recognition Vendor Test 2000
Appendix IV: Facial Recognition Technology Page 177 GAO- 03- 174
Biometrics for Border Security
evaluation used the Face Recognition Technology (FERET) Database, which
DOD*s Counterdrug Technology Development Program Office sponsors. The
evaluation report was issued on February 16, 2001.
The first category, the recognition performance test, evaluated all
algorithms on a standardized database collected by a universal sensor.
Participating vendors were given 72 continuous hours in which to compare
13,872 images to one another, amounting to more than 192 million
comparisons. Three vendors completed this portion: Identix, Viisage, and
C- VIS Computer Vision and Automation GmbH. Banque Technology Systems
International Ltd. (Banque- Tec) and Miros Inc. (E- True Technology), two
other vendors, were able to compare only approximately 4, 000 of the
13,872 images in the allotted time, and their results were not included.
Following this test, different environmental studies were conducted to
show how the system responded to numerous variables such as distance,
lighting, and facial expressions. We describe a sample of the results from
a number of environmental studies, noting the overall lack of appreciable
difference between the match accuracy of the Viisage and Identix
algorithms. For the identification experiments, the charts we present show
the probability that a vendor*s top match correctly identified
individuals. For the verification experiments, the results show the
probability of correct verification while holding the FMR constant at
0.01. Each probe image was taken with a camera and matched by the vendor*s
system to the gallery images, which were drawn from FERET and other large
databases.
Distance Experiments
The distance experiments were designed to evaluate the performance of
face- matching algorithms on images of subjects at different distances
from the fixed camera. For the distance experiments, the probe images were
taken at varying distances and compared, using the vendor*s system, to
gallery images that were taken at a distance of between 1.5 and 2 meters
(see figures 40 and 41).
Appendix IV: Facial Recognition Technology Page 178 GAO- 03- 174
Biometrics for Border Security
Figure 40: Facial Recognition Distance Identification
Source: GAO analysis of FRVT 2000 data. 0.05
0.24 0.33
0.02 0.29
0.18 0.0
0.14 0.06
0 0.5
1.0 Probability of identification
2 35 Camera distance (meters)
C- VIS Viisage Identix 0.05 0.02
0.29 0.18
0 0.14
0.06 0.24
0.33
Appendix IV: Facial Recognition Technology Page 179 GAO- 03- 174
Biometrics for Border Security
Figure 41: Facial Recognition Distance Verification
Source: GAO analysis of FRVT 2000 data.
Across all algorithms, the three sets of distance experiments indicated
that performance decreases as the distance between the person and camera
increases. At a distance of 5 meters, Viisage, the best vendor in this
category, could correctly identify the image only about 13.7 percent of
the time.
Expression Experiments
The expression tests evaluated how well identification and verification
work when comparing images of the same person with different facial
expressions. In this test, the gallery image was a face with a specific
expression, and the probe image was the same face with an alternative
expression. Identification proved more sensitive to change in expression
than verification. Viisage and Identix correctly identified and verified
more than 80 percent of the images (see figures 42 and 43).
Probability of correct verification 2 35 Camera distance (meters)
C- VIS Viisage Identix
0 0.2
0.4 0.6
0.8 1.0
0.08 0.04
0.27 0.21
0.01 0.14
0.05 0.31
0.43
Appendix IV: Facial Recognition Technology Page 180 GAO- 03- 174
Biometrics for Border Security
Figure 42: Facial Recognition Expression Identification
Source: GAO analysis of FRVT 2000 data.
Figure 43: Facial Recognition Expression Verification
Source: GAO analysis of FRVT 2000 data.
0 Probability of identification
C- VIS Viisage Identix 0.5
1.0
0.37 0.82
0.95
0 Probability of correct verification
C- VIS Viisage Identix 0.5
1.0
0.38 0.85
0.96
Appendix IV: Facial Recognition Technology Page 181 GAO- 03- 174
Biometrics for Border Security Media Experiments
The media experiments were designed to evaluate the performance of face-
matching algorithms when comparing images stored on different media. This
application may be useful in comparing older mug shots to newer pictures
taken with digital cameras. For Viisage and Identix, switching between 35
mm gallery images and digital probe images, and vice versa, did not
significantly affect performance (see figures 44 and 45).
Figure 44: Facial Recognition Media Identification: Digital to 35 mm
Source: GAO analysis of FRVT 2000 data.
Probability of identification 0
C- VIS Viisage Identix 0.5
1.0
0.55 0.96 0.98
Appendix IV: Facial Recognition Technology Page 182 GAO- 03- 174
Biometrics for Border Security
Figure 45: Facial Recognition Media Verification: Digital to 35 mm
Source: GAO analysis of FRVT 2000 data.
Pose Experiments
The pose experiments measured the effect of different viewpoints on
identification. They attempted to match a frontal gallery image with probe
images that were rotated various degrees away from the front. The results
reflected the best score of all vendors at each degree. As the degrees
from the frontal image increased, the probability of identification fell
rapidly. At 60 degrees away from the frontal image, identification was
correct only 30 percent of the time (see figure 46).
Probability of correct verification 0
C- VIS Viisage Identix 0.5
1.0
0.53 0.96 0.98
Appendix IV: Facial Recognition Technology Page 183 GAO- 03- 174
Biometrics for Border Security
Figure 46: Facial Recognition Pose Identification
Note: These results reflect the best scores of all vendors at each degree.
Source: FRVT 2000.
Temporal Experiments
Temporal experiments addressed the effect of time delay between a first
and subsequent capture of facial images. The test attempted to match each
probe image with a gallery image of the same person taken approximately 1
year earlier. These experiments showed that a vendor*s ability to
correctly identify and verify images decreases significantly with time.
After 1 year, Viisage and Identix identified 31 percent and 48 percent of
faces, respectively. Viisage correctly verified 41 percent of images,
Identix 56 percent (see figures 47 and 48).
Probability of identification 0
15 25 40 60 Degrees from frontal image 0.5
1.0 0.98 0.94
0.68 0.3
Appendix IV: Facial Recognition Technology Page 184 GAO- 03- 174
Biometrics for Border Security
Figure 47: Facial Recognition Temporal Identification
Note: The time period measured was 1 year. Source: GAO analysis of FRVT
2000 data.
Figure 48: Facial Recognition Temporal Verification
Note: The time period measured was 1 year. Source: GAO analysis of FRVT
2000 data.
0 Probability of identification
C- VIS Viisage Identix 0.5
1.0
0.04 0.31
0.48
0 Probability of correct verification
C- VIS Viisage Identix 0.5
1.0
0.05 0.41
0.56
Appendix IV: Facial Recognition Technology Page 185 GAO- 03- 174
Biometrics for Border Security
The second test category, product usability, evaluated the complete facial
recognition system rather than just the facial recognition algorithm. An
access control scenario with live subjects was chosen. Five vendors
reported results, including the three vendors that completed the
recognition performance test and the two that did not. When discussing the
results, however, it is important to note that some systems tested were
not intended for access control applications.
The two product usability tests were the enrollment timed test (ETT) and
the old image database timed test (OIDTT). The tests had two main
differences: (1) the subjects were stationary for the ETT and walked
toward the camera in the OIDTT and (2) all vendors performed substantially
better on the ETT, in which they enrolled the images under their own
systems, than on the OIDTT, in which the images were provided to them
before the test. Also, the facial recognition systems were quicker and
more accurate in the verification experiments than in the identification
experiments. See table 25 for the results.
Table 25: Facial Recognition Product Usability Test Old image database
timed Enrollment timed Vender Percent
verified Percent identified Percent
verified Percent identified
Banque- Tec 7% 0 22% 22% C- VIS 0 0 69 83 Identix 64 31% 78 52 Miros (E-
True) 36 0 78 71 Viisage 0 0 84 84
Note: Percentages are correct matches. Matching that took longer than 10
seconds counted as failure.
Source: GAO analysis of FRVT 2000 data.
Facial Recognition Vendor Test 2002 (FRVT 2002), a follow- up to FRVT
2000, does not use the FERET database and is not a live facial recognition
test. 2 Since the variables involved with a live capture do not allow for
an equal test bed among all the participants, databases of photograph
images will be used. Also, this test will include video data to determine
whether multiple images of a person increase matching accuracy.
2 Fifteen different agencies are sponsors of FRVT 2002, including the
Defense Advanced Research Projects Agency, the National Institute of
Justice, and the Transportation Security Administration. NIST is selecting
images and computing test scores. Facial Recognition Vendor Test
2002
Appendix IV: Facial Recognition Technology Page 186 GAO- 03- 174
Biometrics for Border Security
Twenty- seven organizations are participating in FRVT 2002, each testing
for a minimum of 4 days and a maximum of 11 days with its own hardware and
software (i. e., its own algorithms). The tests will perform a 100
kilobyte by 100 kilobyte comparison (comparing each face to every other
face in the database) and return the results in the form of similarity
matrixes. In preliminary tests of facial recognition, NIST has seen a 75
percent probability of verification with a 1 percent probability of false
acceptance, compared with fingerprint recognition*s 95 percent probability
of verification and 1 percent probability of false acceptance.
NPL conducted a performance evaluation of seven biometric systems from May
through December 2000, producing a final biometric product testing report
on March 19, 2001. The facial recognition portion of the test used an
Identix FaceIt Verification demonstration as well as alternative
enrollment and matching algorithms.
The 0 percent FTER included persons unable to present the required
biometric feature, those unable to produce an image of sufficient quality
at enrollment, and those unable to reproduce their biometric feature
consistently. At an FMR of about 1 percent, the facial recognition system
with the alternate matching algorithm had an FNMR of approximately 3.3
percent. Additional experimental results were
The facial recognition system collected a sequence of images over a 10-
second period, saving the best match. This resulted in a mean transaction
time of 15 seconds, a median of 14 seconds, and a minimum of 10 seconds.
The matching algorithm could make 800 matches per minute with a Pentium
processor, a Windows interface, and a Windows 2000 operating system. These
diagnostic programs had significant overhead, so the matching algorithm
may be significantly faster than the results showed, perhaps by a factor
exceeding 100.
Tests also found that males had a lower FNMR than females, and the FNMR
for attempts made immediately following enrollment were significantly
lower than those made at a volunteer*s second or third visit.
For a personnel identification application, the Army Research Laboratory
tested an identification system from July through October 2001, using
Identix facial recognition technology. With 270 participants,
approximately 42,000 face identification attempts were made. Despite the
vendor*s claims Biometric Product Testing
U. S. Army Research Laboratory Test
Appendix IV: Facial Recognition Technology Page 187 GAO- 03- 174
Biometrics for Border Security
of a 75 percent rate of correct identification, the testing showed that
only 51 percent were correctly identified. Further, the correct
identification was in the system*s top 10 possible matches only 81 percent
of the time, instead of the 99.3 percent that the vendor claimed.
Inadequate lighting was a primary performance issue.
In 2001, FAA and Safe Skies tested a facial recognition technology system
for access control of airport employees. Following the test, the biometric
system was removed. Twenty- eight people successfully enrolled in an
average of 3 minutes and 2 seconds. The test included operational testing
in a normal environment as well as testing under a controlled environment.
The FNMR for the operational test was approximately 26 percent. Before
device installation, the time required to pass through the door was
approximately 2 seconds; after installation, 11.5 seconds. The FNMR for
the controlled test was approximately 3 percent. Under normal test
conditions, the rate of passage through the door was about six people per
minute.
Test results for abnormal conditions were as follows:
FNMR was nearly 100 percent when test subjects enrolled without
sunglasses but passed through the device with sunglasses. The opposite*
enrolling with sunglasses and presenting with sunglasses* also yielded an
FNMR of nearly 100 percent.
When test subjects enrolled without reading glasses but passed through
the device with reading glasses, FNMR was nearly 60 percent; when they
enrolled with and presented without reading glasses, FNMR was nearly 20
percent.
FNMR increased notably for one test subject of three with a 5- day beard
growth. No effect was noted for the two other subjects. The effect was
little or none for enrolling with 5- day beard growth and then attempting
access while clean- shaven.
A horizontal 3/4- inch adhesive bandage on the chin produced an overall
FNMR of 40 percent, but the results were highly dependent on the test
subject* three had a rate of 0 percent and two had a rate of 100 percent.
A round bandage on the cheek produced an overall FNMR of 6 percent.
No effect was noted from high backlighting directly; however, one test
subject with glasses was falsely rejected 10 of 10 times. Further FAA
Facial Recognition Test
Appendix IV: Facial Recognition Technology Page 188 GAO- 03- 174
Biometrics for Border Security
investigation showed a reflection on the glasses from the backlighting
from the door window.
The Department of State Bureau of Consular Affairs evaluated facial
recognition technology for identifying ineligible visa applicants. Viisage
and Identix provided facial recognition software. The final evaluation
report was issued on January 30, 2001.
Laboratory testing involving data sets of 10,000 to 100,000 images
revealed that less than 30 percent of intentionally seeded duplicate
images were correctly matched. This translates into an FNMR of around 70
percent. The processing speed for facial recognition enrollment was more
than adequate. Images were aligned and enrolled at a rate of approximately
two per second for both tested products. Processing speed for search
ranged from excellent for one vendor*s product to marginal for the other
vendor*s product. In the latter case, an improved version of the software,
submitted after formal testing was completed, was faster by a factor of
two in performing searches of large data sets. The search speed might
limit its usefulness in processing a large data set but is acceptable for
daily operations.
The National Visa Center tested the technology with the diversity visa
program in the field. This trial showed that a facial recognition system
can be successful in identifying matches involving duplicate applications.
More than 500 matches were found while examining more than 5,000 of 35,000
possible duplicate images. Of these 500 and more, 146 represented cases
that had not been discovered by other means. This success was obtained
despite the obviously poor quality of the pictures submitted, the poor
capture characteristics of the Quickcam cameras used, and the less than
optimal scanning technique the data entry personnel used. It was observed
that Identix*s product was more forgiving of the image quality problems
and generally reported more matches.
Despite the vendors* cooperative, responsive, and interactive approach in
supplying testable products and engineering support, the facial
recognition software packages, even in their *final* versions, following
numerous developmental versions, exhibited significantly troublesome
behavior* such as corrupt databases, poorly implemented capabilities, and
the need for workaround solutions* that impeded testing. State Department
Consular
Affairs Tests
Appendix IV: Facial Recognition Technology Page 189 GAO- 03- 174
Biometrics for Border Security
INS conducted a facial verification test for the Secure Electronic Network
for Travelers Rapid Inspection (SENTRI) from November 1997 through July
1998 at California*s Otay Mesa port of entry. The facial verification test
involved taking video images of drivers at an inspection booth. The video
clips were compared to the SENTRI enrollment database of photographs for
all drivers in the SENTRI lane. An Identix system was used for the tests.
The experiment found that pictures taken in a full frontal enrollment pose
showed a significantly higher recognition rate than pictures taken when
the head was rotated slightly. It also found a principal identification
problem when the image was obtained during validation. Obscured faces that
were hidden by part of the vehicle and those with excessive glare or
extreme shadows were essentially unusable. In testing, the proportion of
video clips exhibiting these properties was initially very high. Adding
cameras increased the chance of getting an unobstructed video clip. A new
camera system using fuzzy logic helped reduce glare and shadows.
With these changes, the system was able to get usable images for
approximately 90 percent of the vehicles in a lane. With such images, the
system had an FNMR of 1.6 percent and a low EER of 2.1 percent. The report
concluded that the facial verification system performed admirably in a
challenging environment.
The State Department is conducting pilots using facial recognition
technology from Identix and Viisage to compare images from 23 of its
posts. The facial recognition software is used primarily to compare
digital pictures in one- to- many matching to identify people who apply
more than once for nonimmigrant visas or diversity visas. 3 A secondary
one- to- many matching of photographs from both previously issued visas
and new visa
3 The annual Diversity Visa Lottery Program makes 50,000 immigrant visas
available through a lottery to people who wish to come to the United
States from countries with low immigration rates. Winners are chosen
randomly from all qualified entries by the State Department*s National
Visa Center. Border Control
Applications Piloted and Deployed
INS SENTRI State Department Posts
Appendix IV: Facial Recognition Technology Page 190 GAO- 03- 174
Biometrics for Border Security
applications is performed against a watch list database. The photographs
from all visa applications are scanned into the system, regardless of
whether visas are issued or applications are rejected. All scanned images
(not just the templates) are retained in case future versions of the
facial recognition software use a different template format.
The primary performance factor for the State Department pilots has been
the quality of the photographs submitted with applications. The better the
quality of the photographs is, the more likely it is that match results
will be good. It was found that many of the images in the databases are
poor in quality* either too dark or too light for facial recognition,
poorly focused, or distorted in some other way. Consequently, the State
Department is working to develop standards for photograph quality. Age was
found to be a performance factor. For example, both Identix and Viisage
have found it difficult to match children because their faces change
rapidly. However, State Department officials have not noticed any
appreciable differentiation in the quality between the Identix and Viisage
match algorithms.
Of approximately 197,000 images (applicants* photographs) for diversity
visas processed in the 2002 program year, 75 percent were successfully
enrolled in the diversity visa facial recognition database. The images
from the 74,348 successful applications were matched against the
enrollment database. About 6,000 candidate matches were made; 85 percent
were determined to be actual matches. The facial recognition technology
identified 60 individuals who submitted multiple applications that were
not detected by the manual process.
In October 2001, 23 posts processed approximately 26,000 nonimmigrant visa
images, of which 78 percent were successfully enrolled in the nonimmigrant
visa facial recognition database. For all 23 posts, around 4,000 candidate
matches were made. The percentage of actual matches varied by post, as one
post*s matching had an FMR of 1 percent, and another post*s matching
resulted in an FMR of 65 percent.
One of the first major installations of facial recognition technology at
an airport was at Iceland*s Keflavik International Airport in June 2001.
As a result of Iceland*s participation in Europe*s Schengen agreement,
border controls between that country and others participating in the
agreement Iceland
Appendix IV: Facial Recognition Technology Page 191 GAO- 03- 174
Biometrics for Border Security
have been eliminated. 4 The facial recognition system was implemented to
identify known criminals and false asylum seekers while maintaining a
level of convenience for citizen travelers.
The Basel Project is a pending implementation of facial recognition and
hand geometry for day workers entering and exiting Israel by way of the
Gaza Strip. Fingerprint technology was rejected, since the primary users
are laborers whose fingerprints are unreliable as a biometric for
matching.
Individuals enrolling at the Israeli- Palestinian land border will receive
a contactless smart card with a high- resolution picture and a hand
geometry biometric. As they enter and leave Israel, they will be processed
through 42 routing passages to unattended checkpoints at verification
terminals inside a building. An estimated 60,000 verifications will be
processed daily, performing a one- to- one match against a stored template
in a central server, with a backup stored on the smart card.
Australia*s Sydney Airport is conducting a facial recognition pilot to
determine cost effectiveness and efficiency in an operational environment.
The technology is being used for both verification and identification.
Oneto- one verification is performed to identify false passports as
travelers present their passports, and one- to- many identification is
used to identify terrorists among the crowds.
The Dominican Republic is implementing Identix*s facial recognition
technology for scanning passports at 120 entry points. The system will
capture a face biometric, which will be used in a search against a central
criminal watch list database. If another biometric is needed in the
future, the passport reader will also be capable of reading a fingerprint.
Processing speed for facial recognition enrollment is approximately two
images per second. The raw search speed is one million searches per second
on a single computer, but other factors are involved, such as the
4 The Schengen agreement, begun in 1985, is designed to facilitate travel
within the European Union. Passengers flying between member countries now
leave from domestic rather than international airport terminals,
eliminating the need to present travel documents when entering and
exiting. The Schengen agreement went into effect in Iceland on March 25,
2001. Israel
Australia Dominican Republic Processing Issues
Appendix IV: Facial Recognition Technology Page 192 GAO- 03- 174
Biometrics for Border Security
size of the database. For an identification application, search speed can
be dramatically improved by storing some templates on a disc during
alignment for use during later searches. A facial recognition system can
be designed to achieve a desired response time by increasing the number of
processors, but the trade- off to increased speed is greater cost.
Because facial recognition biometrics can be used in various applications,
different requirements affect performance time differently. The
requirements for performing a background check and a duplicate face check
at enrollment would differ from those for performing verifications at
borders. Verifications at a border would be practically instantaneous if
performing a one- to- one match against a template stored on a travel
document or a smart card, but an additional one- to- many watch list
search would add time, depending on the size of the database. Facial
recognition results in a faster response time than fingerprint recognition
in a one- tomany search. The implication of a heavily queried database is
that a priority level must be assigned to determine when the various
transactions are to be handled.
In surveillance applications, travelers would not interact physically with
the cameras and computers that run the facial recognition technology. The
durability of this equipment would depend on the manufacturer*s
specifications.
Because lighting is such a major performance factor, the use of awnings or
shades with outdoor installations of facial recognition technology could
be required to block direct light. Without awnings or shade, glare or
shadows might present a problem that could be compounded by reflections
from nearby buildings or vehicles. Device Durability and
Environmental Constraints
Appendix V: Iris Recognition Technology Page 193 GAO- 03- 174 Biometrics
for Border Security
Iris recognition technology was developed in 1992 and is therefore one of
the newest of the commercially available biometric technologies. It is
based on the distinct, visible characteristics of the eye*s iris, the
colored ring that surrounds the pupil (see figure 49). Built from elastic
connective tissue, the iris is a very rich source of biometric data. The
characteristics of the iris are formed during the eighth month of
gestation and do not change except through actions such as refractive
surgery, cataract surgery, and cornea transplants. Iris recognition can
even be used to verify the identity of blind people as long as one of
their sightless eyes has an iris.
Figure 49: The Iris and Other Parts of the Eye
Source: Copyright, the American Academy of Ophthalmology.
The iris has more numerous and dense forms of variability than other
biometrics. Whereas traditional biometrics have only 13 to 60 distinct
characteristics, the iris can be said to have 266 unique spots, and iris
recognition technology uses some 173 of these features. The primary
visible characteristic of the iris is the trabecular meshwork, tissue that
gives the appearance of dividing the iris radially. Other features include
striations, rings, furrows, a corona, and freckles.
Besides the iris* many distinctive characteristics, its patterns also
differ substantially from person to person. A person*s left and right eyes
have different iris patterns, and the irises of identical twins have
almost no statistical similarity. It has been postulated that the
probability of two persons having the same iris pattern is 1 in 7 billion.
Appendix V: Iris Recognition Technology
Appendix V: Iris Recognition Technology Page 194 GAO- 03- 174 Biometrics
for Border Security
An iris recognition system uses a small high- quality camera to capture a
black- and- white, high- resolution picture of the iris. The technology
relies on infrared imaging, using wavelengths from 700 to 900 nanometers,
a range the American Academy of Ophthalmology has stated is safe.
How close the person should be to the camera and her level of
participation depend on the type of system. Physical access control
applications require a person to stand within 3 to 10 inches of the camera
and center the iris in a mirror within an area 1 inch square directly in
front of the camera (see figure 50). The system may prompt the person to
move slightly forward or backward to allow a proper image capture. Systems
using desktop cameras to control logical access to computers and networks
require a distance of about 18 inches to capture the iris image (see
figure 51). Users must center their eyes on the camera with a guidance
light or hologram. Personal identification systems, such as those at
airport kiosks in trusted traveler applications, allow users to stand as
far away as 3 feet. However, users must remain still as the camera locates
the eye and captures the image.
Figure 50: Iris Recognition Physical Access Control System
Source: Panasonic Digital Communications & Security Co.
How the Technology Works
Appendix V: Iris Recognition Technology Page 195 GAO- 03- 174 Biometrics
for Border Security
Figure 51: Iris Recognition System with Desktop Camera
Source: Panasonic Digital Communications & Security Co.
An iris recognition system first defines the boundaries of the iris,
establishes a coordinate system over the iris, and defines the zones for
analysis within the coordinate system. Feature extraction algorithms map
the segments of the iris into hundreds of independent vectors that define
the orientation and spatial frequency of the distinctive features, along
with the position of the features. However, the entire iris is not used: A
portion of the top as well as 45 degrees of the bottom remain unused, to
account for pupil dilation, occlusion from eyelids, and reflection from
the camera (see figure 52).
Appendix V: Iris Recognition Technology Page 196 GAO- 03- 174 Biometrics
for Border Security
Figure 52: Mapping the Eye for Iris Recognition Systems
Source: Dr. John Daugman, Cambridge University, Cambridge, U. K.
Algorithms also check for the presence of a pattern on the sphere of the
eye instead of on an internal plane and use measurements at different
wavelengths to determine that the eye is living. The visible
characteristics within the zones are then converted into a 512 byte
template that is used to identify or verify the identity of an individual;
256 of these bytes contain control information.
Iridian Technologies Inc. is the sole owner and developer of iris
recognition technology. Iridian markets applications through hardware
manufacturers and systems integrators, including Argus Solutions,
EyeTicket Corp., IBM, Joh. Enschede Security Solutions, LG Electronics,
NEC Singapore, Oki Electric Industry Co., Panasonic, SAFLINK Corp.,
Siemens AG, Titan Corp., and Unisys.
Iris recognition systems cost approximately $2,000 for physical access
units. The camera itself costs $200.
Some users are unable to provide adequate enrollment images because they
find the iris image capture process too difficult. Poor eyesight may also
hinder the ability of some people to line up their eyes with the camera.
Colored and bifocal contact lenses can affect system performance, and so
can exceptionally strong glasses. People with The Leading Vendors
The Cost of Devices Performance Issues
Appendix V: Iris Recognition Technology Page 197 GAO- 03- 174 Biometrics
for Border Security
glaucoma may not be reliably identified. Also, glare and reflections,
along with user settling and distraction, can cause interference.
Some people resist technologies that scan the eye, but unlike biometric
identification and verification technologies such as fingerprint
recognition or hand geometry, iris recognition technology requires no body
contact. Iris recognition technology is more user friendly than retina
recognition systems in that no light source is shone into the eye and
close proximity to the scanner is not required. However, iris recognition
does use active infrared illumination in the 700 to 900 nanometer
wavelength range. It has none of the inherent risks associated with
lasers. Some people assume that the imaging of their irises will reveal
their medical data, such as heart disease, diabetes, and high blood
pressure, but images of the iris acquired for iris recognition reveal no
information about a person*s health.
Iris recognition is being used operationally for physical access control,
logical access control, and personal identification applications. An
EyeTicket access control system was installed at Douglas International
Airport in Charlotte, North Carolina, in July 2000 to control airline and
airport employee access to restricted areas. The company has also
installed the access control system at Germany*s Frankfurt Airport.
Iridian has installed IrisAccess* at Baltimore Technologies* data hosting
center in Sydney, Australia. Access to the highly secure facility requires
that anyone requesting entry verify her identity with both a proximity
card and the iris recognition technology.
The Office of Legislative Counsel for the U. S. House of Representatives
has recently installed an iris recognition system to protect confidential
computer files and working documents. Iris recognition systems have been
deployed in several prison systems in the United States to prevent inmates
from swapping identities with visitors as well as to verify the identity
of prisoners before they are released. User Acceptance
The Technology*s Maturity
Operational Uses
Appendix V: Iris Recognition Technology Page 198 GAO- 03- 174 Biometrics
for Border Security
NPL conducted a performance evaluation of seven biometric systems from May
through December 2000. The iris portion used Iridian*s IriScan System
2200. The FTER was 0.5 percent. The FMR was 0 percent and the iris
recognition system had an FNMR of 1.9 percent. Additional experimental
results were that
the iris system had a mean transaction time of 12 seconds, a median of
10 seconds, and a minimum of 4 seconds;
the matching algorithm could make 1.5 million matches per minute when
using a SunUltra5 with a SunOS 5.8 operating system, 270 MHz processor,
and 128 Mb of memory; and
people without glasses had a lower FNMR than those with glasses. The U.
S. Army Research Laboratory recently tested an Iridian verification
system. There were 186,918 eye identification attempts on 93,459
registrations. The FMR was well below 1 percent. Despite the vendor*s
claims of greater than 99.5 percent correct identification, the testing
showed a 6 percent FNMR; glare and reflections appeared to be primary
culprits in this discrepancy. User settling and distraction also
contributed to the problem.
In April 1996, Sandia National Laboratories evaluated a prototype
biometric recognition system provided by IriScan. Average enrollment time
was 2 minutes and 15 seconds. During the first phase of the test, there
was a raw FNMR of 11.8 percent. After removing the errors that could be
attributed to extreme environmental conditions or deliberate misuse, the
FNMR became 10.2 percent. The average transaction time of a sampling of
transactions was 14 seconds. The minimum transaction time recorded was 6
seconds, the maximum 23 seconds. Users attempted 96 false match
transactions with no actual false matches. Overall, the researchers
concluded that the system performed extremely well in difficult
conditions.
Researchers at c*t Magazine in Germany set out to see whether they could
fool Panasonic*s Authenticam BM- ET100, a desktop iris recognition system.
The investigators* first attempts to spoof the system by using iris images
projected on monitors failed because of the too intense reflection of
light. However, they succeeded in beating the system by holding up to Test
s
Biometric Product Testing U. S. Army Research Laboratory Sandia National
Laboratories
c*t Magazine
Appendix V: Iris Recognition Technology Page 199 GAO- 03- 174 Biometrics
for Border Security
the camera a high- resolution picture of an iris with a tiny hole cut out
to allow the pupil of a live eye shine through. They also found it
possible to enroll with the aid of this artificial eye. From that point
on, anyone in possession of the eye pattern was able to log on to the
system. Moreover, the system also matched the iris of the person whose
picture had been used to create the artificial eye with the enrolled
reference template.
Iris recognition has been used in some border control environments. For
example, beginning in July 2001, frequent travelers on transatlantic
Virgin Atlantic Airways and British Airways flights have been able to
bypass passport control at London*s Heathrow Airport, without waiting in
line for an immigration agent. In trial runs, 2,000 American and Canadian
passengers have undergone identity checks by British immigration officers
before being enrolled. Once registered and enrolled, they can proceed, as
arriving passengers, directly to specific lanes to verify their identity
against a biometric template stored in a central database (see figures 53
and 54). If the verification is successful, they are issued a ticket
admitting them directly to the United Kingdom. The trial is being operated
by the airlines and involves no changes to passports. Border Control
Applications Piloted and Deployed
United Kingdom
Appendix V: Iris Recognition Technology Page 200 GAO- 03- 174 Biometrics
for Border Security
Figure 53: Iris Recognition Device for Border Control at London*s Heathrow
Airport
Source: EyeTicket Corporation.
Figure 54: Border Control Lane with Iris Recognition Device at London*s
Heathrow Airport
Source: EyeTicket Corporation.
Appendix V: Iris Recognition Technology Page 201 GAO- 03- 174 Biometrics
for Border Security
The Canada Customs and Revenue Agency has initiated the Expedited
Passenger Processing System, which will include iris recognition
technology. The system will allow frequent travelers to expedite
inspection. It is planned to be operational at Lester B. Pearson
International Airport in Toronto and Vancouver International Airport at
the beginning of 2003. An enrollment of about 200,000 spread out over 5
years is expected. The plan is to use a central database for storing the
iris templates. Initially, it was not clear whether computer performance
would allow for a central database, so provision was made for a token to
store the biometric. However, testing has shown that doing the checks
centrally does not significantly affect performance time. Either one- to-
one matches (with an identifying token) will be made or one- to- many,
with the system identifying applicants by the iris match.
In October 2001, an iris recognition system was installed at Amsterdam*s
Schiphol Airport. The system expedites the way for travelers from 18
European countries into the Netherlands and includes about 2,000 frequent
travelers. Users must go through a two- phase process. First, passengers
undergo a background check, a passport review, and an iris scan. The
template is encrypted and embedded on a smart card. This phase takes about
15 minutes. The second phase identifies and verifies each registered
traveler at the immigration checkpoint. The traveler*s reference template
is compared with a real- time scan of the iris. This process typically
takes about 10 to 15 seconds and allows the passenger to bypass long
immigration lines. The Schiphol program charges each enrolled traveler a
yearly fee of $89 to use the system. The FNMR is less than 1 percent; the
FMR is less than 0.001 percent.
Iris recognition is used to admit workers who travel into Singapore from
Malaysia each day by motorcycle. The workers* irises are scanned by a
camera installed in kiosks in designated lanes, instead of their having to
present their paperwork to an official. About 50,000 workers cross the
border each day.
In February 2002, at the King Abdul Aziz Airport in Jeddah, Saudi Arabia,
iris recognition tracked and identified visitors who were on pilgrimage
for the Hajj season of worship. The process included a random check at
passport control, enrollment into a database, and subsequent
identification on departure. The systems were in place to ensure that
visitors did not overstay their visas and also to identify potential
security threats. It is estimated that images of 20,000 to 30,000 irises
were collected. Canada
Netherlands Singapore Saudi Arabia
Appendix V: Iris Recognition Technology Page 202 GAO- 03- 174 Biometrics
for Border Security
Although iris recognition systems can perform both one- to- many
identification and one- to- one verification, they are deployed primarily
for identification. In some processors, iris recognition technology can
search hundreds of thousands of records per second. Very few biometrics
have the capability of iris recognition for a high- speed exhaustive
search of a database.
Because iris recognition systems use infrared illumination, they can be
used in the dark. Their durability depends greatly on the specifications
of the system*s individual components. Processing Issues
Device Durability and Environmental Constraints
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 203 GAO- 03- 174 Biometrics for Border Security
For each of the four scenarios, we created cost models to estimate the
cost of developing, implementing, and maintaining various biometric
systems. Besides including the cost of purchasing the biometric hardware,
we estimated costs for additional hardware, software, maintenance,
personnel, training, and effects on other procedures in order to derive
lifecycle cost estimates. We followed the cost element structure that DOD
uses at acquisition program milestone and decision reviews to assess major
automated information systems costs. Tailoring this structure to reflect
our four scenarios, we used it to standardize costs so that they could be
compared at a high level. We present the costs in two parts. Initial costs
represent the costs required to plan, design, develop, and field the
system. Recurring costs represent the annual costs required to operate and
continually maintain the system to keep it in operation.
We estimated seven sets of initial cost elements: costs for systems
engineering and program management; development, installation, and
training; biometric hardware; biometric software; network infrastructure;
renovating consular facilities; and hardware infrastructure upgrades.
Systems engineering and program management costs included both program
management activities and government in- house engineering efforts to
design, develop, and test the biometric system. For the watch list
scenarios, we used an engineering build- up of personnel and their
respective costs. For issuing visas and passports with biometrics, we used
an overall factor of the total initial cost to estimate this effort.
Development, installation, and training costs included all resources
required to design, develop, test, and implement a biometric system. For
the watch list scenarios, we used an analogy to the Consular Lookout and
Support System (CLASS) to estimate the cost of developing and implementing
a watch list database. For issuing visas and passports with biometrics, we
used an analogy to IAFIS and applied an engineering scaling factor to
account for additional biometric storage space.
Biometric hardware costs included costs for biometric scanners, token card
readers, and token cards for storing biometric data as well as costs for
the personal computers to make these devices function properly. To
estimate costs, we used average vendor costs where available and, in other
cases, we relied on expert opinion.
Biometric software costs included the licensing cost for biometric
scanners, card readers, and database software. For the watch list Appendix
VI: Cost Estimates for Using
Biometrics for Border Security Initial Cost Elements
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 204 GAO- 03- 174 Biometrics for Border Security
scenarios, we used cost estimates provided by the State Department, based
on analogy to CLASS. For issuing visas and passports with biometrics, we
assumed this cost was already included in the development cost for IAFIS.
Network infrastructure included costs associated with purchasing and
installing the local area networks needed to establish the connectivity
required by the biometric systems. For the watch list scenarios, we used
cost estimates provided by the State Department, based on an analogy to
CLASS. For issuing visas and passports with biometrics, we used an analogy
to a trusted traveler cost estimate developed by IBG.
To issue visas with biometrics, additional space at the consulates and
embassies will be required to accommodate the new process of capturing
applicants* biometrics. For the watch list scenarios, the consular
facility cost is for the renovation of primary and contingency space for
the new computer systems. We used square foot data provided by the State
Department to estimate this cost. We did not include costs for the
collection of biometrics at passport acceptance offices because most of
these are not State Department facilities, and we had no basis on which to
estimate the appropriate amount of space for these offices.
Hardware infrastructure upgrades included the cost to refresh hardware
every 3 years. To estimate this element, we calculated the cost to replace
one- third of the hardware annually, an accepted industry standard and the
practice for the State Department*s visa and passport sites.
We estimated 10 sets of recurring cost elements: program management,
biometric hardware maintenance, software and system maintenance, network
infrastructure maintenance, consular operating personnel, port of entry
operating personnel, communications, training, consular facility
maintenance, and annual supplies.
Program management included the cost of providing continuing program
management over the system*s useful life. To estimate this cost for the
watch list scenarios, we used an engineering build- up of personnel and
their respective costs. For issuing visas and passports with biometrics,
we estimated this cost to be 20 percent of the initial systems engineering
and program management cost.
Biometric hardware maintenance included the cost of providing maintenance
and repair for the biometric and system hardware. We used Recurring Cost
Elements
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 205 GAO- 03- 174 Biometrics for Border Security
an average factor of 12.5 percent, based on a 10 percent to 15 percent
range IBG provided in its trusted traveler cost estimate.
Software and system maintenance costs included annual software licensing
for databases plus costs for personnel to upgrade and maintain them. For
the watch list scenarios, we used an engineering build- up of personnel
and their respective costs. For issuing visas and passports with
biometrics, we used an analogy to IAFIS annual system costs, applying the
engineering scaling factor to account for additional database storage of
the various biometrics.
Network infrastructure maintenance included the cost of providing hardware
and software maintenance for the network. For the watch list scenarios, we
used data from the State Department, based on its experience from CLASS.
For issuing visas and passports with biometrics, we used the same factor
of 12.5 percent that was used for estimating hardware maintenance.
The costs for consular operating personnel are for visa operating
personnel at embassies and consulates around the world or for passport
operating personnel at passport acceptance offices. For the checking of a
biometric watch list before issuing visa, we estimated that one additional
staff member per embassy or consulate would be required to resolve watch
list hits. We did not include additional staff for checking a biometric
watch list before issuing a passport. For the issuance of visas with
biometrics, we first estimated the number of personnel needed at the
consulates, using time to capture the biometrics as a variable. We then
estimated the cost for the foreign service nationals who would perform the
capturing, the foreign service officers who would oversee them, and
auxiliary consulate staff to assist during peak load times. The annual
costs for all visa operating personnel and the one- time moving costs for
new foreign service nationals and officers were provided to us by the
State Department. For the issuance of passports with biometrics, we
assumed one staff member per passport acceptance office to troubleshoot
problems with the biometric equipment.
Port of entry operating personnel include staff to troubleshoot biometrics
at ports of entry. To estimate costs for these personnel, we made the
assumption that there would be three staff per port of entry who would be
trained and able to troubleshoot problems arising from biometric capturing
or the inability to match biometric data.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 206 GAO- 03- 174 Biometrics for Border Security
The costs of communications included the cost of maintaining a wide area
network able to provide secure electronic connectivity from the consular
and port of entry sites to a headquarters location for comparing
biometrics. To estimate this element, we used an analogy to IAFIS
communication costs with a cost- per- location methodology.
Training included the costs to train personnel in using biometrics,
including the cost of travel. We used an average of $5,000 per staff
annually to estimate this cost.
The cost of maintaining consular facilities included maintaining newly
acquired space. We used data on cost per square foot provided by the State
Department.
In estimating the cost of annual supplies, we included the cost to
purchase biometric token cards for the storage of biometrics collected for
issuing passports and visas. This cost also includes the amortized cost of
the infrastructure required to produce the cards, including elements such
as centralized certificate issuance servers, key management components,
and the card management infrastructure. We used data provided by the State
Department for the Mexican border crossing card.
We prepared the life- cycle cost estimates using fiscal year 2002 constant
dollars* that is, inflation was not considered for the multiple years over
which funds would be required for acquisition* and they represent rough
order of magnitude costs. Following are the assumptions that frame the
boundary of our cost estimates.
Scenario life- cycle cost estimates represent development and
installation time plus 10 years* operational life.
Phasing of costs over time is simplified, and actual schedules to both
develop and install equipment and infrastructure will most likely differ.
Biometric technologies* fingerprint, facial, and iris recognition*
represent standardization to a single vendor*s protocols. Biometric
technology costs represent the average costs of vendors* products. Four
flat fingerprints will be collected for fingerprint recognition.
There are 210 visa- issuing embassies and consulates worldwide. There
are 4,500 passport acceptance offices. There are 3,950 primary and
secondary inspection stations at 400 ports of entry. Assumptions
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 207 GAO- 03- 174 Biometrics for Border Security
Personnel costs reflect both direct costs and indirect costs. Three
personnel will be needed to troubleshoot equipment at ports of entry, or
1,200 additional staff.
No costs were estimated for
additional inspectors at ports of entry,
additional facility space for passport acceptance offices or at ports of
entry for primary and secondary inspections,
biometric equipment for exiting the United States, and
biometric security technology (e. g., encryption of biometric data). We
used the following assumptions to create the cost estimates for the two
biometric watch list scenarios:
The watch list database will include 10 million records.
Matches will be performed using facial recognition technology.
To conduct watch list checks before issuing travel documents, facial
images will be generated by capturing the physical photographs applicants
present when they apply for a visa or passport.
The images will be collected and scanned at consulates and embassies for
visas and at passport acceptance offices and transmitted through
telecommunications resources to a central facility in metropolitan
Washington, D. C.
The estimated costs for conducting biometric watch list checks before
travelers are issued travel documents and before they enter the country
are shown in table 26. Estimated Costs for
Conducting Watch List Checks with Biometrics
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 208 GAO- 03- 174 Biometrics for Border Security
Table 26: Estimated Costs for Watch List Checks before Issuing Travel
Documents and before Entering the United States Watch list check before
issuing travel documents
Watch list check before entering the United States a
Cost element Initial cost Annual recurring cost Initial cost Annual
recurring cost Investment
Systems engineering and program management $540 $540 Development;
installation; training 7,900 207,900 Initial biometric hardware 6,045
16,488 Initial biometric software 4, 600 4,600 Network infrastructure 100
100,000 Consular facility renovation 570 570 Hardware infrastructure
upgrade $2,523 $38,969
Operations and support
Program management 540 540 Biometric hardware maintenance 926 14,761
Software and system maintenance 6, 400 38,560 Network infrastructure
maintenance 100 12,513 Visa operating personnel 33,075 50,715 Port of
entry operating personnel 94,679 Communications 10,540 10,038 Recurring
training 1,000 26,750 Consular facility maintenance 152 152
Total $52,830 $72,896 $330,197 $236,960
Note: In thousands of fiscal year 2002 constant dollars. a Numbers do not
sum because of rounding.
Source: GAO analysis.
We developed cost estimates for six different combinations of biometric
technologies under two different possibilities for issuing visas. The
State Department receives about 10.3 million visa applications each year.
In fiscal year 2000, INS estimated that approximately 14 million
individuals traveled under the visa waiver program. If these travelers
must obtain a visa to travel to the United States, we assume that this
same number would also be required to have their biometric sample
collected. We used the following assumptions to estimate the costs of
adding biometrics to visas:
The number of visa applicants will remain constant at 10.3 million
annually. The number of travelers in the visa waiver program will remain
constant at 14 million annually. Estimated Costs for
Issuing Visas with Biometrics
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 209 GAO- 03- 174 Biometrics for Border Security
Enrolling travelers using a single biometric (whether for fingerprint,
facial, or iris recognition) is estimated at 6 minutes (10 applicants
enrolled per hour).
Enrolling travelers using multiple biometrics (for example, fingerprint
and facial combined, fingerprint and iris combined, or fingerprint,
facial, and iris combined) is estimated at 10 minutes (6 applicants
enrolled per hour).
All current visa- issuing embassies and consulates will be equipped to
collect biometrics from visa applicants.
Costs were not included for additional inspectors or facility space at
ports of entry. Tables 27* 32 show the cost of issuing visas with
biometrics using fingerprint recognition, iris recognition, facial
recognition, fingerprint and iris recognition, fingerprint and facial
recognition, and fingerprint, iris, and facial recognition.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 210 GAO- 03- 174 Biometrics for Border Security
Table 27: Estimated Costs for Issuing Visas with Biometrics Using
Fingerprint Recognition Annual visa applicants 10.3 million with visa
waiver program 24.3 million
without visa waiver program Cost element Initial cost Annual
recurring cost Initial cost Annual recurring cost
Investment costs
Systems engineering and program management $111,147 $145,645 Development;
installation; training 527,655 558,936 Initial biometric hardware 219,033
443,241 Initial biometric software Network infrastructure 152,500 152,500
Consular facility renovation 335,781 463,606 Hardware infrastructure
upgrade $79,114 $93,986
Operations and support
Program management 22,229 29,129 Biometric hardware maintenance 10,905
16,538 Software and system maintenance 73,123 125,292 Network
infrastructure maintenance 19,063 19,063 Visa operating personnel 75,926
111,626 114,903 150,603 Port of entry operating personnel 94,679 94,679
Communications 20,577 20,577 Recurring training 32,472 38,040 Consular
facility maintenance 89,541 123,628 Annual supplies (cards) 154,809
365,229
Total $1,422,042 $708,138 $1,878,832 $1,076,765
Note: In thousands of fiscal year 2002 constant dollars. Numbers do not
sum because of rounding. Source: GAO analysis.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 211 GAO- 03- 174 Biometrics for Border Security
Table 28: Estimated Costs for Issuing Visas with Biometrics Using Iris
Recognition Annual visa applicants 10.3 million with visa waiver program
24.3 million
without visa waiver program Cost element Initial cost Annual
recurring cost Initial cost Annual recurring cost
Investment
Systems engineering and program management $110,925 $145,375 Development;
installation; training 527,655 558,936 Initial biometric hardware 216,563
440,240 Initial biometric software Network infrastructure 152,500 152,500
Consular facility renovation 335,781 463,606 Hardware infrastructure
upgrade $78,298 $92,996
Operations and support
Program management 22,185 29,075 Biometric hardware maintenance 10,596
16,163 Software and system maintenance 73,123 125,292 Network
infrastructure maintenance 19,063 19,063 Visa operating personnel 75,926
111,626 114,903 150,603 Port of entry operating personnel 94,679 94,679
Communications 20,577 20,577 Recurring training 32,472 38,040 Consular
facility maintenance 89,541 123,628 Annual supplies (cards) 154,809
365,229
Total $1,419,349 $706,970 $1,875,562 $1,075,346
Note: In thousands of fiscal year 2002 constant dollars. Numbers do not
sum because of rounding. Source: GAO analysis.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 212 GAO- 03- 174 Biometrics for Border Security
Table 29: Estimated Costs for Issuing Visas with Biometrics Using Facial
Recognition Annual visa applicants 10.3 million with visa waiver program
24.3 million
without visa waiver program Cost element Initial cost Annual
recurring cost Initial cost Annual recurring cost
Investment
Systems engineering and program management $109,258 $143,350 Development;
installation; training 527,655 558,936 Initial biometric hardware 198,037
417,737 Initial biometric software Network infrastructure 152,500 152,500
Consular facility renovation 335,781 463,606 Hardware infrastructure
upgrade $72,185 $85,570
Operations and support
Program management 21,852 28,670 Biometric hardware maintenance 8,280
13,350 Software and system maintenance 73,123 125,292 Network
infrastructure maintenance 19,063 19,063 Visa operating personnel 75,926
111,626 114,903 150,603 Port of entry operating personnel 94,679 94,679
Communications 20,577 20,577 Recurring training 32,472 38,040 Consular
facility maintenance 89,541 123,628 Annual supplies (cards) 154,809
365,229
Total $1,399,156 $698,207 $1,851,033 $1,064,702
Note: In thousands of fiscal year 2002 constant dollars. Numbers do not
sum because of rounding. Source: GAO analysis.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 213 GAO- 03- 174 Biometrics for Border Security
Table 30: Estimated Costs for Issuing Visas with Biometrics Using
Fingerprint and Iris Recognition Annual visa applicants 10.3 million with
visa waiver program 24.3 million
without visa waiver program Cost element Initial cost Annual
recurring cost Initial cost Annual recurring cost
Investment
Systems engineering and program management $151,218 $193,935 Development;
installation; training 820,165 867,087 Initial biometric hardware 253,098
495,336 Initial biometric software Network infrastructure 228,750 228,750
Consular facility renovation 378,188 563,655 Hardware infrastructure
upgrade $119,315 $145,299
Operations and support
Program management 30,244 38,787 Biometric hardware maintenance 16,601
26,444 Software and system maintenance 96,591 176,331 Network
infrastructure maintenance 28,594 28,594 Visa operating personnel 95,044
130,744 160,006 195,706 Port of entry operating personnel 94,679 94,679
Communications 20,577 20,577 Recurring training 70,405 88,966 Consular
facility maintenance 100,850 150,308 Annual supplies (cards) 154,809
365,229
Total $1,926,463 $863,409 $2,508,769 $1,330,920
Note: In thousands of fiscal year 2002 constant dollars. Source: GAO
analysis.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 214 GAO- 03- 174 Biometrics for Border Security
Table 31: Estimated Costs for Issuing Visas with Biometrics Using
Fingerprint and Facial Recognition Annual visa applicants 10.3 million
with visa waiver program 24.3 million
without visa waiver program Cost element Initial cost Annual
recurring cost Initial cost Annual recurring cost
Investment
Systems engineering and program management $149,375 $191,495 Development;
installation; training 820,165 867,087 Initial biometric hardware 232,621
468,231 Initial biometric software Network infrastructure 228,750 228,750
Consular facility renovation 378,188 563,655 Hardware infrastructure
upgrade $112,557 $136,354
Operations and support
Program management 29,875 38,299 Biometric hardware maintenance 14,042
23,056 Software and system maintenance 96,591 176,331 Network
infrastructure maintenance 28,594 28,594 Visa operating personnel 95,044
130,744 160,006 195,706 Port of entry operating personnel 94,679 94,679
Communications 20,577 20,577 Recurring training 70,405 88,966 Consular
facility maintenance 100,850 150,308 Annual supplies (cards) 154,809
365,229
Total $1,904,143 $853,723 $2,479,223 $1,318,099
Note: In thousands of fiscal year 2002 constant dollars. Numbers do not
sum because of rounding. Source: GAO analysis.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 215 GAO- 03- 174 Biometrics for Border Security
Table 32: Estimated Costs for Issuing Visas with Biometrics Using
Fingerprint, Iris, and Facial Recognition Annual visa applicants 10.3
million with visa waiver program 24.3 million
without visa waiver program Cost element Initial cost Annual
recurring cost Initial cost Annual recurring cost
Investment
Systems engineering and program mnagement $177,371 $221,694 Development;
installation; training 1,027,676 1,090,238 Initial biometric hardware
259,924 504,372 Initial biometric software Network infrastructure 305,000
305,000 Consular facility renovation 378,188 563,655 Hardware
infrastructure upgrade $150,527 $182,402
Operations and support
Program management 35,474 44,339 Biometric hardware maintenance 18,893
30,967 Software and system maintenance 119,661 226,432 Network
infrastructure maintenance 38,125 38,125 Visa operating personnel 95,044
130,744 160,006 195,706 Port of entry operating personnel 94,679 94,679
Communications 20,577 20,577 Recurring training 105,608 133,449 Consular
facility maintenance 100,850 150,308 Annual supplies (cards) 154,809
365,229
Total $2,243,202 $969,947 $2,844,964 $1,482,212
Note: In thousands of fiscal year 2002 constant dollars. Numbers do not
sum because of rounding. Source: GAO analysis.
We used the following assumptions to estimate the costs of adding
biometrics to passports:
The number of passport applicants will remain constant at 7 million
annually.
Enrolling travelers using a single biometric (whether for fingerprint,
facial, or iris recognition) is estimated at 6 minutes (10 applicants
enrolled per hour).
Enrolling travelers using multiple biometrics (for example, fingerprint
and facial combined, fingerprint and iris combined, or fingerprint,
Estimated Costs for
Issuing Passports with Biometrics
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 216 GAO- 03- 174 Biometrics for Border Security
facial, and iris combined) is estimated at 10 minutes (6 applicants
enrolled per hour).
All current passport acceptance offices will be equipped to collect
biometrics passport applicants.
Costs were not included for additional inspectors or facility space at
ports of entry. Costs are also not included for additional facility space
at passport acceptance offices.
Tables 33* 38 show the cost of issuing passports with biometrics using
fingerprint recognition, iris recognition, facial recognition, fingerprint
and iris recognition, fingerprint and facial recognition, and fingerprint,
iris, and facial recognition.
Table 33: Estimated Costs for Issuing Passports with Biometrics Using
Fingerprint Recognition
Cost element Initial cost Annual recurring cost
Investment
Systems engineering and program management $370,797 Development;
installation; training 2,665,282 Initial biometric hardware 229,685
Initial biometric software Network infrastructure 1,225,000 Consular
facility renovation Hardware infrastructure upgrade $450,488
Operations and support
Program management 74,159 Biometric hardware maintenance 17,514 Software
and system maintenance 58,146 Network infrastructure maintenance 153,125
Passport operating personnel 443,805 Port of entry operating personnel
94,679 Communications 122,962 Recurring training 53,875 Consular facility
maintenance Annual supplies (cards) 105,210
Total $4,490,764 $1,573,965
Note: In thousands of fiscal year 2002 constant dollars. Numbers do not
sum because of rounding. Source: GAO analysis.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 217 GAO- 03- 174 Biometrics for Border Security
Table 34: Estimated Costs for Issuing Passports with Biometrics Using Iris
Recognition
Cost element Initial cost Annual recurring cost
Investment
Systems engineering and program management $370,366 Development;
installation training 2,665,282 Initial biometric hardware 224,898 Initial
biometric software Network infrastructure 1,225,000 Consular facility
renovation Hardware infrastructure upgrade $448,908
Operations and support
Program management 74,073 Biometric hardware maintenance 16,916 Software
and system maintenance 58,146 Network infrastructure maintenance 153,125
Passport operating personnel 443,805 Port of entry operating personnel
94,679 Communications 122,962 Recurring training 53,875 Consular facility
maintenance Annual supplies (cards) 105,210
Total $4,485,545 $1,571,700
Note: In thousands of fiscal year 2002 constant dollars. Numbers do not
sum because of rounding. Source: GAO analysis.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 218 GAO- 03- 174 Biometrics for Border Security
Table 35: Estimated Costs for Issuing Passports with Biometrics Using
Facial Recognition
Cost element Initial cost Annual recurring cost
Investment
Systems engineering and program management $367,135 Development;
installation; training 2,665,282 Initial biometric hardware 188,991
Initial biometric software Network infrastructure 1,225,000 Consular
facility renovation Hardware infrastructure upgrade $437,059
Operations and support
Program management 73,427 Biometric hardware maintenance 12,428 Software
and system maintenance 58,146 Network infrastructure maintenance 153,125
Passport operating personnel 443,805 Port of entry operating personnel
94,679 Communications 122,962 Recurring training 53,875 Consular facility
maintenance Annual supplies (cards) 105,210
Total $4,446,407 $1,554,716
Note: In thousands of fiscal year 2002 constant dollars. Numbers do not
sum because of rounding. Source: GAO analysis.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 219 GAO- 03- 174 Biometrics for Border Security
Table 36: Estimated Costs for Issuing Passports with Biometrics Using
Fingerprint and Iris Recognition
Cost element Initial cost Annual recurring cost
Investment
Systems engineering and program management $552,750 Development;
installation; training 4,026,605 Initial biometric hardware 277,560
Initial biometric software Network infrastructure 1,837,500 Consular
facility renovation Hardware infrastructure upgrade $670,993
Operations and support
Program management 110,550 Biometric hardware maintenance 24,476 Software
and system maintenance 67,777 Network infrastructure maintenance 229,688
Passport operating personnel 443,805 Port of entry operating personnel
94,679 Communications 122,962 Recurring training 107,750 Consular facility
maintenance Annual supplies (cards) 105,210
Total $6,694,415 $1,977,890
Note: In thousands of fiscal year 2002 constant dollars. Source: GAO
analysis.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 220 GAO- 03- 174 Biometrics for Border Security
Table 37: Estimated Costs for Issuing Passports with Biometrics Using
Fingerprint and Facial Recognition
Cost element Initial cost Annual recurring cost
Investment
Systems engineering and program management $549,518 Development;
installation; training 4,026,605 Initial biometric hardware 241,654
Initial biometric software Network infrastructure 1,837,500 Consular
facility renovation Hardware infrastructure upgrade $659,144
Operations and support
Program management 109,904 Biometric hardware maintenance 19,988 Software
and system maintenance 67,777 Network infrastructure maintenance 229,688
Passport operating personnel 443,805 Port of entry operating personnel
94,679 Communications 122,962 Recurring training 107,750 Consular facility
maintenance Annual supplies (cards) 105,210
Total $6,655,277 $1,960,906
Note: In thousands of fiscal year 2002 constant dollars. Numbers do not
sum because of rounding. Source: GAO analysis.
Appendix VI: Cost Estimates for Using Biometrics for Border Security
Page 221 GAO- 03- 174 Biometrics for Border Security
Table 38: Estimated Costs for Issuing Passports with Biometrics Using
Fingerprint, Iris, and Facial Recognition
Cost element Initial cost Annual recurring cost
Investment
Systems engineering and program management $723,821 Development;
installation; training 5,302,929 Initial biometric hardware 289,529
Initial biometric software Network infrastructure 2,450,000 Consular
facility renovation Hardware infrastructure upgrade $879,648
Operations and support
Program management 144,764 Biometric hardware maintenance 26,950 Software
and system maintenance 77,407 Network infrastructure maintenance 306,250
Passport operating personnel 443,807 Port of entry operating personnel
94,679 Communications 122,962 Recurring training 161,625 Consular facility
maintenance Annual supplies (cards) 105,210
Total $8,766,279 $2,363,302
Note: In thousands of fiscal year 2002 constant dollars. Source: GAO
analysis.
Appendix VII: Comments from the U. S. Department of State
Page 222 GAO- 03- 174 Biometrics for Border Security
Appendix VII: Comments from the U. S. Department of State
Appendix VII: Comments from the U. S. Department of State
Page 223 GAO- 03- 174 Biometrics for Border Security
Appendix VIII: Comments from the U. S. Department of Justice
Page 224 GAO- 03- 174 Biometrics for Border Security
Appendix VIII: Comments from the U. S. Department of Justice
Appendix VIII: Comments from the U. S. Department of Justice
Page 225 GAO- 03- 174 Biometrics for Border Security
Appendix VIII: Comments from the U. S. Department of Justice
Page 226 GAO- 03- 174 Biometrics for Border Security
Appendix VIII: Comments from the U. S. Department of Justice
Page 227 GAO- 03- 174 Biometrics for Border Security
Page 228 GAO- 03- 174 Biometrics for Border Security
Appendix IX: GAO Contacts and Acknowledgments
Page 229 GAO- 03- 174 Biometrics for Border Secur
Nancy R. Kingsbury (202) 512- 2700; kingsburyn@ gao. gov. Naba Barkakati
(202) 512- 4499; barkakatin@ gao. gov.
Additional staff who made major contributions to this report were
Venkareddy Chennareddy, Barbara Hills, Ashfaq Huda, Richard Hung,
Elizabeth Johnston, John C. Martin, Eric Ow, Madhav Panwar, Penny Pickett,
Tracy Pierson, David Plocher, and Karen Richey.
We gratefully acknowledge the time and assistance of the following people
who reviewed a draft of this report: Dennis Carlton, International
Biometric Group; Paul Collier, The Biometric Foundation; Larry Hornak,
West Virginia University; Anil Jain, Michigan State University; Rick
Lazarick, Transportation Security Administration; Peter Neumann, SRI
International; Lee Tien, Electronic Frontier Foundation; Jim Wayman, San
Jose State University; Charles Wilson, National Institute of Standards and
Technology; and John Woodward, RAND Corporation.
We also appreciate the contributions provided by the following
organizations during our meetings on biometrics and border security:
Airports Council International; American Civil Liberties Union; American
Immigration Lawyers Association; Biometric Technology Inc.; Border Trade
Alliance; Cameron County Bridge Systems; Cogent Systems Inc.; Electronic
Data Systems Corp.; Electronic Privacy Information Center; EyeTicket
Corp.; Graphco Technologies Inc.; Identix Inc.; International Biometric
Industry Association; International Organization of Masters, Mates, and
Pilots; Iridian Technologies Inc.; Mitretek Systems Inc.; National Council
La Raza; Recognition Systems Inc.; Sagem Morpho Inc.; and Viisage
Technology Inc. Appendix IX: GAO Contacts and
Acknowledgments GAO Contacts Acknowledgments
Bibliography Page 230 GAO- 03- 174 Biometrics for Border Security
American Association of Motor Vehicle Administrators. AAMVA National
Standard for the Driver License/ Identification Card. AAMVA DL/ ID- 2000.
June 30, 2000. http:// www. aamva. org/ standards/
American National Standards Institute. Biometric Information and Security.
ANSI X9.84- 2001, 2001. http:// www. ansi. org/ default. asp
BioAPI Consortium. BioAPI Specification Version 1.1, March 16, 2001.
http:// www. bioapi. org/
Blackburn, Duane M., Mike Bone, and P. Jonathon Phillips. Facial
Recognition Vendor Test 2000: Evaluation Report. Sponsored by DOD
Counterdrug Technology Development Program Office, Defense Advanced
Research Projects Agency, and National Institute of Justice, February 16,
2001. http:// www. frvt. org/ FRVT2000/ documents. htm
International Civil Aviation Organization. Machine Readable Travel
Documents Technical Report: Selection of a Globally Interoperable
Biometric for Machine- Assisted Identity Confirmation with MRTDS, 1st ed.,
2001. http:// www. icao. int/ index. cfm
Jain, Anil, Ruud Bolle, and Sharath Pankanti, eds. Biometrics: Personal
Identification in Networked Society. Boston: Kluwer Academic Publishers,
1999.
Maio, Dario, and others. *FVC 2000: Fingerprint Verification Competition,*
technical report, University of Bologna, Department of Electronics,
Computer Science, and Systems, September 2000.
Mansfield, Tony, and others. Biometric Product Testing Final Report.
Middlesex, Eng.: National Physical Laboratory, March 19, 2001. Nanavati,
Samir, Michael Thieme, and Raj Nanavati. Biometrics: Identity Verification
in a Networked World. New York: John Wiley & Sons, 2002.
National Institute of Standards and Technology. Common Biometric Exchange
File Format (CBEFF), NISTIR 6529. Gaithersburg, Md.: January 3, 2001.
Phillips, P. Jonathon, and others. *An Introduction to Evaluating
Biometric Systems.* IEEE Computer 33: 2 (February 2000): 56* 63.
Bibliography
Bibliography Page 231 GAO- 03- 174 Biometrics for Border Security
U. S. General Accounting Office. Computer Security: Improvements Needed to
Reduce Risk to Critical Federal Operations and Assets. GAO02- 231T.
Washington, D. C.: November 9, 2001.
U. S. General Accounting Office. Electronic Benefits Transfer: Use of
Biometrics to Deter Fraud in the Nationwide EBT Program. GAO/ OSI- 95- 20.
Washington, D. C.: September 29, 1995.
U. S. General Accounting Office. Federal Funding for Selected Surveillance
Technologies. GAO- 02- 438R. Washington, D. C.: March 14, 2002.
U. S. General Accounting. Homeland Security: Challenges and Strategies in
Addressing Short- and Long- Term National Needs. GAO- 02- 160T.
Washington, D. C.: November 7, 2001.
U. S. General Accounting Office. Identity Fraud: Prevalence and Links to
Alien Illegal Activities. GAO- 02- 830T. Washington, D. C.: June 25, 2002.
U. S. General Accounting Office. Information Security: Advances and
Remaining Challenges to Adoption of Public Key Infrastructure Technology.
GAO- 01- 277. Washington, D. C.: February 26, 2001.
U. S. General Accounting Office. Information Security: Opportunities for
Improved OMB Oversight of Agency Practices. GAO/ AIMD- 96- 110.
Washington, D. C.: September 24, 1996.
U. S. General Accounting Office. National Preparedness: Technologies to
Secure Federal Buildings. GAO- 02- 687T. Washington, D. C.: April 25,
2002.
U. S. General Accounting Office. Social Security Numbers: Government
Benefits from SSN Use but Could Provide Better Safeguards. GAO- 02- 352.
Washington, D. C.: May 31, 2002.
(460525)
The General Accounting Office, the investigative arm of Congress, exists
to support Congress in meeting its constitutional responsibilities and to
help improve the performance and accountability of the federal government
for the American people. GAO examines the use of public funds; evaluates
federal programs and policies; and provides analyses, recommendations, and
other assistance to help Congress make informed oversight, policy, and
funding decisions. GAO*s commitment to good government is reflected in its
core values of accountability, integrity, and reliability.
The fastest and easiest way to obtain copies of GAO documents at no cost
is through the Internet. GAO*s Web site (www. gao. gov) contains abstracts
and fulltext files of current reports and testimony and an expanding
archive of older products. The Web site features a search engine to help
you locate documents using key words and phrases. You can print these
documents in their entirety, including charts and other graphics.
Each day, GAO issues a list of newly released reports, testimony, and
correspondence. GAO posts this list, known as *Today*s Reports,* on its
Web site daily. The list contains links to the full- text document files.
To have GAO e- mail this list to you every afternoon, go to www. gao. gov
and select *Subscribe to daily E- mail alert for newly released products*
under the GAO Reports heading.
The first copy of each printed report is free. Additional copies are $2
each. A check or money order should be made out to the Superintendent of
Documents. GAO also accepts VISA and Mastercard. Orders for 100 or more
copies mailed to a single address are discounted 25 percent. Orders should
be sent to:
U. S. General Accounting Office 441 G Street NW, Room LM Washington, D. C.
20548
To order by Phone: Voice: (202) 512- 6000 TDD: (202) 512- 2537 Fax: (202)
512- 6061
Contact: Web site: www. gao. gov/ fraudnet/ fraudnet. htm E- mail:
fraudnet@ gao. gov Automated answering system: (800) 424- 5454 or (202)
512- 7470
Jeff Nelligan, managing director, NelliganJ@ gao. gov (202) 512- 4800 U.
S. General Accounting Office, 441 G Street NW, Room 7149 Washington, D. C.
20548 GAO*s Mission
Obtaining Copies of GAO Reports and Testimony
Order by Mail or Phone To Report Fraud, Waste, and Abuse in Federal
Programs
Public Affairs
*** End of document. ***