[House Hearing, 108 Congress]
[From the U.S. Government Publishing Office]
EXPLORING COMMON CRITERIA: CAN IT ASSURE THAT THE FEDERAL GOVERNMENT
GETS NEEDED SECURITY IN SOFTWARE?
=======================================================================
HEARING
before the
SUBCOMMITTEE ON TECHNOLOGY, INFORMATION
POLICY, INTERGOVERNMENTAL RELATIONS AND
THE CENSUS
of the
COMMITTEE ON
GOVERNMENT REFORM
HOUSE OF REPRESENTATIVES
ONE HUNDRED EIGHTH CONGRESS
FIRST SESSION
__________
SEPTEMBER 17, 2003
__________
Serial No. 108-126
__________
Printed for the use of the Committee on Government Reform
Available via the World Wide Web: http://www.gpo.gov/congress/house
http://www.house.gov/reform
______
92-771 U.S. GOVERNMENT PRINTING OFFICE
WASHINGTON : 2003
____________________________________________________________________________
For Sale by the Superintendent of Documents, U.S. Government Printing Office
Internet: bookstore.gpo.gov Phone: toll free (866) 512-1800; (202) 512�091800
Fax: (202) 512�092250 Mail: Stop SSOP, Washington, DC 20402�090001
COMMITTEE ON GOVERNMENT REFORM
TOM DAVIS, Virginia, Chairman
DAN BURTON, Indiana HENRY A. WAXMAN, California
CHRISTOPHER SHAYS, Connecticut TOM LANTOS, California
ILEANA ROS-LEHTINEN, Florida MAJOR R. OWENS, New York
JOHN M. McHUGH, New York EDOLPHUS TOWNS, New York
JOHN L. MICA, Florida PAUL E. KANJORSKI, Pennsylvania
MARK E. SOUDER, Indiana CAROLYN B. MALONEY, New York
STEVEN C. LaTOURETTE, Ohio ELIJAH E. CUMMINGS, Maryland
DOUG OSE, California DENNIS J. KUCINICH, Ohio
RON LEWIS, Kentucky DANNY K. DAVIS, Illinois
JO ANN DAVIS, Virginia JOHN F. TIERNEY, Massachusetts
TODD RUSSELL PLATTS, Pennsylvania WM. LACY CLAY, Missouri
CHRIS CANNON, Utah DIANE E. WATSON, California
ADAM H. PUTNAM, Florida STEPHEN F. LYNCH, Massachusetts
EDWARD L. SCHROCK, Virginia CHRIS VAN HOLLEN, Maryland
JOHN J. DUNCAN, Jr., Tennessee LINDA T. SANCHEZ, California
JOHN SULLIVAN, Oklahoma C.A. ``DUTCH'' RUPPERSBERGER,
NATHAN DEAL, Georgia Maryland
CANDICE S. MILLER, Michigan ELEANOR HOLMES NORTON, District of
TIM MURPHY, Pennsylvania Columbia
MICHAEL R. TURNER, Ohio JIM COOPER, Tennessee
JOHN R. CARTER, Texas CHRIS BELL, Texas
WILLIAM J. JANKLOW, South Dakota ------
MARSHA BLACKBURN, Tennessee BERNARD SANDERS, Vermont
(Independent)
Peter Sirh, Staff Director
Melissa Wojciak, Deputy Staff Director
Rob Borden, Parliamentarian
Teresa Austin, Chief Clerk
Philip M. Schiliro, Minority Staff Director
Subcommittee on Technology, Information Policy, Intergovernmental
Relations and the Census
ADAM H. PUTNAM, Florida, Chairman
CANDICE S. MILLER, Michigan WM. LACY CLAY, Missouri
DOUG OSE, California DIANE E. WATSON, California
TIM MURPHY, Pennsylvania STEPHEN F. LYNCH, Massachusetts
MICHAEL R. TURNER, Ohio
Ex Officio
TOM DAVIS, Virginia HENRY A. WAXMAN, California
Bob Dix, Staff Director
Chip Walker, Professional Staff Member
Ursula Wojciechowski, Clerk
David McMillen, Minority Professional Staff Member
C O N T E N T S
----------
Page
Hearing held on September 17, 2003............................... 1
Statement of:
Davidson, Mary Ann, chief security officer, Server Technology
Platforms, Oracle.......................................... 73
Fleming, Michael G., Chief, Information Assurance Solutions,
Information Assurance Directorate, National Security Agency 21
Gorrie, Robert G., Deputy Director, Defensewide Information
Assurance Program Office, Office of the Assistant Secretary
of Defense for Networks and Information Integration, and
DOD Chief Information Officer.............................. 43
Klaus, Christopher W., chief technology officer, Internet
Security Systems, Inc...................................... 83
Roback, Edward, Chief, Computer Security Division, National
Institute of Standards and Technology, U.S. Department of
Commerce................................................... 7
Spafford, Eugene H., professor and director, Center for
Education and Research in Information Assurance and
Security, Purdue University................................ 88
Thompson, J. David, director, Security Evaluation Laboratory,
Cygnacom Solutions......................................... 66
Letters, statements, etc., submitted for the record by:
Clay, Hon. Wm. Lacy, a Representative in Congress from the
State of Missouri, prepared statement of................... 56
Davidson, Mary Ann, chief security officer, Server Technology
Platforms, Oracle, prepared statement of................... 76
Fleming, Michael G., Chief, Information Assurance Solutions,
Information Assurance Directorate, National Security
Agency, prepared statement of.............................. 23
Gorrie, Robert G., Deputy Director, Defensewide Information
Assurance Program Office, Office of the Assistant Secretary
of Defense for Networks and Information Integration, and
DOD Chief Information Officer, prepared statement of....... 46
Klaus, Christopher W., chief technology officer, Internet
Security Systems, Inc., prepared statement of.............. 85
Putnam, Hon. Adam H., a Representative in Congress from the
State of Florida, prepared statement of.................... 4
Roback, Edward, Chief, Computer Security Division, National
Institute of Standards and Technology, U.S. Department of
Commerce, prepared statement of............................ 10
Spafford, Eugene H., professor and director, Center for
Education and Research in Information Assurance and
Security, Purdue University, prepared statement of......... 90
Thompson, J. David, director, Security Evaluation Laboratory,
Cygnacom Solutions, prepared statement of.................. 69
EXPLORING COMMON CRITERIA: CAN IT ASSURE THAT THE FEDERAL GOVERNMENT
GETS NEEDED SECURITY IN SOFTWARE?
----------
WEDNESDAY, SEPTEMBER 17, 2003
House of Representatives,
Subcommittee on Technology, Information Policy,
Intergovernmental Relations and the Census,
Committee on Government Reform,
Washington, DC.
The subcommittee met, pursuant to notice, at 10:10 a.m., in
room 2154, Rayburn House Office Building, Hon. Adam Putnam
(chairman of the subcommittee) presiding.
Present: Representatives Putnam, Clay and Watson.
Staff present: Bob Dix, staff director; John Hambel, senior
counsel; Chip Walker, professional staff; Ursula Wojciechowski,
clerk; Suzanne Lightman, fellow; Erik Glavich, legislative
assistant; Ryan Hornbeck, intern; David McMillen, minority
professional staff member; and Jean Gosa, minority chief clerk.
Mr. Putnam. The Subcommittee on Technology, Information
Policy, Intergovernmental Relations and the Census will come to
order. Good morning, and I apologize for running a few minutes
late. I have 20 high school students in Washington for a week
for a congressional classroom program to become familiar with
the city and our government and how everything works. None of
us were figuring on Hurricane Isabel, so we are trying to
figure out a way to get 20 airline tickets in very short order,
and it's not going to be terribly easy.
Welcome to another important hearing on cybersecurity.
Today the subcommittee continues its aggressive oversight and
examination of the information security issues most important
to our Nation. As many of you know, Secretary Ridge announced
the creation of the U.S. Computer Emergency Response Team
[U.S.-CERT] in conjunction with Carnegie Mellon University.
This is an important step in the progress that needs to be made
by our government in protecting the Nation's computers from
cyber attack. It's no longer a question of if our computer
networks will be attacked, but rather when, how often and to
what degree.
Experts from the government and the private sector who have
come before this subcommittee are very concerned that the
United States is not adequately prepared to ward off a serious
cyber attack that could cause severe economic devastation as
well as contribute potentially to the loss of life. Blaster and
SoBigF are stark examples of how worm and virus vulnerabilities
can cost us billions of dollars in lost productivity and
administrative costs in a very short period of time. From the
home user to the private enterprise to the Federal Government,
we all need to take the cyber threat more seriously and move
expeditiously to secure our Nation's computers. I look forward
to continuing to work with the Department of Homeland Security
and other key Federal agencies in this national security
endeavor.
Today's hearing will examine the Common Criteria and
whether or not a similar certification should be applied to all
government software purchasers. For years countries around the
globe have wrestled with the inability to have a commonly
recognized method for evaluating security software. Out of this
climate, the Common Criteria evolved and represents standards
that are broadly useful within the international community.
The international members of the Common Criteria share the
following objectives: to ensure that evaluations of information
technology products and protection profiles are performed to
high and consistent standards, and are seen to contribute
significantly to confidence in the security of those products;
to improve the availability of evaluated, security-enhanced IT
products; to eliminate the burden of duplicating evaluations;
and to continuously improve the efficiency and cost-
effectiveness of the evaluation and certification/validation
process.
The Common Criteria are maintained by an international
coalition and is designed to be useful within the widely
diverse international community. Currently the recognition
arrangement has 15 member countries. The National Security
Agency and NIST represent the United States. Each member
country accepts certificates issued by the members, making the
Common Criteria a global standard. The criteria are technology-
neutral and are designed to be applied to a wide variety of
technologies and levels of security.
The criteria work by providing standardized language and
definitions of IT security components. That standardization
allows the consumer, in our case the Department of Defense, to
create a customized set of requirements for the security of a
product, or protection profile. This profile would include the
level of security assurance that the customer desires,
including the various mechanisms that must be present for
achieving that assurance. Alternatively the criteria allows the
producer of the technology to develop their own set of targets
called a security target. An independent lab overseen by the
participating agencies, in the United States' case NIST and
NSA, then test the product against either the profile or the
target and certifies that it can satisfy the requirements.
Currently the Department of Defense requires Common
Criteria certification for all security-related software
purchases. NSA requires Common Criteria certification for all
purchases for systems classified as national intelligence.
One of the more useful aspects of the Common Criteria is
its ability to allow the purchaser of security software to
compare apples to apples. The protection profile which is cast
in the language of the Common Criteria provides a view of
security features independent of vendor claims. It allows the
purchaser to find out with certainty the security features in a
product and to compare that product with other similar ones to
determine which ones to purchase.
The certification process, conducted by independent labs
overseen by NIST in the United States, concentrates on
analyzing the documentation provided by the vendor testing the
product, documenting the result and reporting it out to its
oversight agency. That agency then reviews the validation
report and issues certification. The process is paid for by the
vendor and can be both expensive and time-consuming. Estimates
for operating systems can be anywhere from 1 to 5 years and
costs in the millions of dollars.
The expense and time commitment of the process has given
rise to some questioning about the usefulness of the process.
For example, the adoption of the Common Criteria could shut
small vendors out of the acquisition process because they might
not have the resources to go through certification. Another
potential problem is the timing. Because certification takes a
significant amount of time, the government might not get the
most cutting-edge technology available. Conversely, the
government does need to gain assurance that security features
in products exist and function as advertised.
This is the larger question that we are faced with: How can
we--governmentwide--get the most secure products available in a
timely and cost-efficient manner and at the same time have IT
companies compete on a level playing field in a competitive
market that rewards rather than stifles innovation? I look
forward to the expert testimony we have assembled today, and I
thank the witnesses for their participation.
As with all of our hearings, today's hearing can be viewed
live via WebCast by going to reform.house.gov. We will hold off
on the other opening statements until the Members arrive, and I
would ask that all of our witnesses comply with the light and
the timing. Your written statement will be submitted for the
record and will be included in its entirety, but we ask that
you summarize your verbal comments to 5 minutes.
[The prepared statement of Hon. Adam H. Putnam follows:]
[GRAPHIC] [TIFF OMITTED] T2771.001
[GRAPHIC] [TIFF OMITTED] T2771.002
[GRAPHIC] [TIFF OMITTED] T2771.003
Mr. Putnam. With that, as is the custom of this
subcommittee, we will swear in the witnesses. I will ask our
first panel rise and raise your right hands.
[Witnesses sworn.]
Mr. Putnam. Note for the record all of the witnesses
responded in the affirmative. And we will move right to our
distinguished panel.
Our first witness is Edward Roback. Mr. Roback serves as
the Chief of the Computer Security Division at the National
Institute of Standards and Technology supporting the agency's
responsibility's to protect sensitive Federal information and
promote security and commercial information technology
products. As Chief, he leads the implementation of NIST
responsibilities under FISMA and Cybersecurity Research and
Development Act. Mr. Roback heads NIST's participation on the
NIST-NSA Technical Working Group and serves on the Committee of
National Security Systems. He has chaired the Federal Agency
Computer Security Programs Managers Forum and co-authored An
Introduction to Computer Security, The NIST handbook. He has
also recently authored NIST's Guidelines to Federal
Organizations on Security Assurance and Acquisition/Use of
Tested/Evaluated Products. For those of who you would like a
copy, they will be available at Barnes and Noble afterwards,
and he will be happy to autograph them for you.
Mr. Roback, you are recognized for 5 minutes. Welcome to
the subcommittee.
STATEMENT OF EDWARD ROBACK, CHIEF, COMPUTER SECURITY DIVISION,
NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY, U.S. DEPARTMENT
OF COMMERCE
Mr. Roback. Thank you, Chairman Putnam. Thank you for the
opportunity to testify today. In response to your invitation, I
first would like to discuss what security assurance is and the
role it plays in overall cybersecurity. I then would like to
turn to the role that security testing and particularly the
Common Criteria and the NIST-NSA-NIAP program play. I would
like to leave you with some ideas as to what else the research
community can do to improve the trust and confidence that we
must have in the proper, correct and secure functioning of
information systems. So let me start.
What is security assurance? If we look at assurance
broadly, it's the basis we need for overall trust and
confidence in the correct and secure information systems. The
overall question of assurance tries to address two questions:
Does the system do what it is supposed to do, and does it not
do the unintended? Within that context, security assurance,
simply put, is the degree of confidence one has that the
security mechanisms of a system is intended. It is not an
absolute guarantee that security is achieved.
How do we get security assurance? There is no single way.
One can get some degree by looking at how a system is built,
the past use of a system, manufacturers' warranties or lack
thereof, and, of course, independent testing and evaluation.
This testing can vary from the straightforward and repeatable
through the more complex and time-consuming. When we have a
standard specification that is very precise, such as with an
encryption algorithm, testing is straightforward, although not
necessarily easy. When a specification is exact, the test can
be correspondingly precise.
On the other hand, when we look at more complex and diverse
IT products which lack common standards specification at the
bits and bytes level, we're often confronted with products
containing millions of lines of code for which a standard spec
does not exist, and testing is not just straightforward.
Testing such products necessarily involves human subjectivity.
NIST refers to such testing as evaluation. NIAP is such a
testing program.
Turning to the NIAP and the CC, in my written statement I
have provided a summary of the development of each, the Mutual
Recognition Arrangement and some of the uses of the criteria
both domestically and overseas, and indeed there have been some
very significant uses. Major issuers of bank cards have formed
work groups to use the Common Criteria to develop a profile for
smart cards; the Financial Services Business Roundtable is
doing that for the financial services community. The Process
Control Security Requirements Forum is using the Common
Criteria for SCADA systems security, and it is also being used
in the health care community trying to use the Common Criteria
to define requirements for the health care systems.
But I think it's important to take a minute to review the
meaning of a Common Criteria certificate. A Common Criteria
evaluation is a measure of the information technology's
compliance to the vendor's claimed security. It is not a
measure or a guarantee that the product is free from malicious
code or that the overall comprised system is secure. Any
product that has a Common Criteria specification can undergo an
evaluation and receive its certificate if the evaluation
process is completed. I provided additional details in my
written statement.
As you mentioned, we have issued advice to the agencies on
the use of evaluated products for non-national security
systems. We described the overall role that assurance can play.
And, of course, the Committee for National Security Systems has
issued its Policy No. 11, and I will defer to my colleagues for
additional comments on that.
As to whether that policy should be extended, I believe
that more data is needed from the CNSS policy experience before
extension is considered or recommended for unclassified
systems. One of the criticisms often levied on NIAP is that
evaluations take too long and cost too much. We hear this from
the small business community. Of course, one would expect to
hear that of any evaluation process that is not free and
instantaneous. However, these products do involve millions of
lines of code. But given resolve, flexibility, resources and
research, significant progress can be made.
For example, the research community should look at new ways
to develop enhanced security testing. We need new methods. The
current process we have is too expensive and involves too much
human subjectivity. We need to invest more in doing such
research, because the sooner we do, the sooner we will have
benefits from the results. We need to look outward at system-
level composability issues and enterprise architecture issues,
and we need to look inward to some of the security issues that
are present with things like protocols. You have to look across
the entire spectrum.
In summary, the Common Criteria provides the means to
develop specifications and a common means to develop security
evaluations. However, more can be done to streamline this
process through research and standards development, resources
permitting. We must also keep in mind that technology alone
will not achieve security, although we are focused on
technology today.
Thank you for the opportunity to testify today.
Mr. Putnam. Thank you very much.
[The prepared statement of Mr. Roback follows:]
[GRAPHIC] [TIFF OMITTED] T2771.004
[GRAPHIC] [TIFF OMITTED] T2771.005
[GRAPHIC] [TIFF OMITTED] T2771.006
[GRAPHIC] [TIFF OMITTED] T2771.007
[GRAPHIC] [TIFF OMITTED] T2771.008
[GRAPHIC] [TIFF OMITTED] T2771.009
[GRAPHIC] [TIFF OMITTED] T2771.010
[GRAPHIC] [TIFF OMITTED] T2771.011
[GRAPHIC] [TIFF OMITTED] T2771.012
[GRAPHIC] [TIFF OMITTED] T2771.013
[GRAPHIC] [TIFF OMITTED] T2771.014
Mr. Putnam. Our second witness is Michael Fleming. Mr.
Fleming currently leads the National Security Agency group
responsible for development and customer implementation support
of a broad set of IA solutions. Prior to this assignment, he
held positions as the Deputy Chief of Network Security Group,
Chief of Network Security Systems Engineering Office, Chief of
Network Security Products Office and special technology
transfer assignment with the NSA Deputy Director For Plans and
Policy. Early in his NSA career Mr. Fleming served in a variety
of technical and program management assignments in
communications security and signals intelligence.
He is a recipient of the NSA Meritorious Civilian Service
Award and twice received the Presidential Rank Award for
Meritorious Service.
It is a pleasure to have you, and you're recognized for 5
minutes.
STATEMENT OF MICHAEL G. FLEMING, CHIEF, INFORMATION ASSURANCE
SOLUTIONS, INFORMATION ASSURANCE DIRECTORATE, NATIONAL SECURITY
AGENCY
Mr. Fleming. Thank you for your interest in cybersecurity,
information security, or information assurance. We have three
words that describe this very important endeavor. I would like
to provide a quick overview of the Common Criteria and the NIAP
current status, some of the potential for even greater
applications, and discuss some of the issues that you have
already raised.
As you stated, it establishes a language which is very
important, a syntax. The criteria is, in fact, a language, a
dictionary for describing user needs and vendor claims. It also
establishes the methodology to make those comparisons in terms
of how well those claims meet needs.
I think it's important to note the criteria does not apply
to all information technology products that make no information
assurance claims. The criteria employs a distinct but related
set of functional requirements which describe the mechanisms
and the assurance requirements which Mr. Roback described in
terms of gaining confidence that those mechanisms work
correctly. There are seven assurance levels, one being the
lowest and the least rigorous; seven being the highest and most
rigorous.
In 1997, we entered a partnership with NIST called NIAP to
promote, demand investment in security products, and establish
the commercial security evaluation capability. To support the
demand the Committee on National Security Systems in January
issued NSTISSP 11, which stipulated the acquisition of
commercial IA products, and IA-enabled products would be
limited to those evaluated under formal schemes such as NIAP.
In terms of demand, we have defined, in fact, 21 protection
profiles, and 31 more are in development. These profiles
address key technology such as operating systems, firewalls,
and intrusion detection systems and other things. And the
demand trend there is encouraging.
As profiles are introduced for a technology, the number of
evaluation claims is increasing. For example, all the operating
systems in evaluation or that have been evaluated are
compliant. All public key infrastructure are compliant. And
about half of the firewall intrusion detection systems are
either claiming compliance or compliant with protection
profile.
As far as the second goal of NIAP, establishing the labs, 8
labs are accredited and have completed 38 evaluations with an
additional 55 underway and more being negotiated continuously.
In terms of expanding the use across a broader spectrum of
environments than just the Department of Defense, the
requirements for information assurance in the national security
market are almost identical to those in other mission-critical
government or commercial systems. Common Criteria can be
leveraged to converge these markets. The larger market would
result in greater return on investment for the vendors, and
everyone in the buying sector would benefit from that leverage.
Regarding limitations, you address cost and timeliness.
Evaluation of timeliness and cost actually is a function of a
number of factors: Product complexity, assurance level aspired
to, the vendor's preparedness to undergo an evaluation. And any
problems found during evaluation typically want to be fixed
before the evaluation is completed. All those can lead to some
time and cost. But a vendor can capitalize on an initial
evaluation investment by reusing parts for subsequent
evaluations for subsequent releases.
While the criteria makes every attempt to identify and
correct security vulnerabilities to ensure--there is no
assurance that these products are bulletproof, especially at
the lower assurance level. Vulnerabilities can be introduced in
a number of ways, from poor design, inappropriate operation.
Source code evaluation is not always required, particularly
until you get to the higher assurance levels, which many
vendors don't aspire to. And vulnerabilities in an IA-enabled
product introduced by unevaluated nonsecurity functionality may
go undetected. Mechanisms complementary to the Common Criteria
are needed to increase our ability to find and eliminate
malicious code in large software applications.
In conclusion, information systems require assurance that
it was specified and designed properly, that it was
independently evaluated against a prescribed set of security
standards, that it will maintain proper operation during its
lifetime even in the face of malicious attacks and human error.
The Common Criteria in NIAP are working. The trends are up, and
process improvements continue. A converged market for security
products would benefit all potential IA buying sectors.
The Common Criteria and NIAP are not a panacea for all
security issues and all information technology. We need
complementary activities. Security needs to be baked into
information systems starting with specification. It cannot just
be evaluated in at the end nor sprinkled in after a system is
fielded. And I think this is an important point in terms of
improving the overall process. This is all about making sure
that a security product is, in fact, secure and doing its job.
It has certainly been my pleasure to discuss the Common
Criteria and share the work of the NIAP with the subcommittee,
and I thank you for the opportunity.
Mr. Putnam. Thank you very much, Mr. Fleming.
[The prepared statement of Mr. Fleming follows:]
[GRAPHIC] [TIFF OMITTED] T2771.015
[GRAPHIC] [TIFF OMITTED] T2771.016
[GRAPHIC] [TIFF OMITTED] T2771.017
[GRAPHIC] [TIFF OMITTED] T2771.018
[GRAPHIC] [TIFF OMITTED] T2771.019
[GRAPHIC] [TIFF OMITTED] T2771.020
[GRAPHIC] [TIFF OMITTED] T2771.021
[GRAPHIC] [TIFF OMITTED] T2771.022
[GRAPHIC] [TIFF OMITTED] T2771.023
[GRAPHIC] [TIFF OMITTED] T2771.024
[GRAPHIC] [TIFF OMITTED] T2771.025
[GRAPHIC] [TIFF OMITTED] T2771.026
[GRAPHIC] [TIFF OMITTED] T2771.027
[GRAPHIC] [TIFF OMITTED] T2771.028
[GRAPHIC] [TIFF OMITTED] T2771.029
[GRAPHIC] [TIFF OMITTED] T2771.030
[GRAPHIC] [TIFF OMITTED] T2771.031
[GRAPHIC] [TIFF OMITTED] T2771.032
[GRAPHIC] [TIFF OMITTED] T2771.033
[GRAPHIC] [TIFF OMITTED] T2771.034
Mr. Putnam. Our third witness for this first panel is
Robert Gorrie. Mr. Gorrie is the National Security Agency
integree serving as the Deputy Director of the Defensewide
Information Assurance Program [DIAP], office, Office of the
Assistant Secretary of Defense for Networks and Information
Integration. Prior to his retirement from the Army after a 26-
year career as a signal officer, he was Chief of the
Information Assurance Division on the Joint Staff and Deputy
Chief of NSA's Information Security Customer Support Office.
Following his retirement, he is employed with Titan Systems
Corp. as vice president of operations in its managed IT
securities service group. He is a graduate of Gannon College
and Penn State University in Pennsylvania as well as both the
Naval and Air War Colleges.
Welcome to the subcommittee. You're recognized.
STATEMENT OF ROBERT G. GORRIE, DEPUTY DIRECTOR, DEFENSEWIDE
INFORMATION ASSURANCE PROGRAM OFFICE, OFFICE OF THE ASSISTANT
SECRETARY OF DEFENSE FOR NETWORKS AND INFORMATION INTEGRATION,
AND DOD CHIEF INFORMATION OFFICER
Mr. Gorrie. Thank you, sir, and I am honored to be here and
pleased to have the opportunity to speak with your committee
about some of the efforts DOD has initiated with respect to the
evaluation of information assurance and information assurance-
enabled products.
As demonstrated in recent operations, U.S. forces have been
extremely successful in the battlefield. They have been able to
translate IT into combat power. However, as our dependence on
IT increases, it creates new vulnerabilities as adversaries
develop new ways of attacking and disrupting U.S. forces.
No one technology operation or person is capable of
protecting the Department's vast networks. In October last
year, the Department published its capstone information
assurance policy. The policy establishes responsibilities and
prescribes procedures for applying integrated, layered
protection for DOD information systems and networks.
The DOD's IA strategies and policies are central to the
committee's Common Criteria question. As I stated, no one
single person, technology or operation can assure DOD's vast
global networks. The Common Criteria, the NIAP evaluation
program, the national and DOD policy addressing IA evaluations
and the evaluated products themselves are part of an integrated
DOD IA strategy.
Even with the solid defense-in-depth strategy in place, we
must be confident in the security and trustworthiness of the
products we use to implement that strategy. New vulnerabilities
in the equipment we use are identified daily. Through the
Department's IA Vulnerability Alert [IAVA], process, users are
made aware of the vulnerabilities and associated fixes. The
IAVA process serves us well, minimizing the effects of recent
cyber incidents on DOD networks. The IAVA process has also
highlighted the alarming rise in the number of vulnerabilities,
the risks they represent and the cost of associated
remediation. Although we continue to improve the efficiency and
effectiveness of the IAVA process, unless we can take proactive
measures to reduce the number of vulnerabilities in our systems
and networks, our ability to respond will begin to degrade.
Although no product will ever be totally secure, we can
incorporate security into the design and, through testing, gain
a reasonable sense of the risk we assume when we use them.
However, that requires policy, enforcement and practice. The
Committee on National Security Systems, their National
Information Assurance Acquisition Policy directs the
acquisition of all COTS IA and IA-enabled products to be used
in national security systems be limited to those which have
been evaluated. Our DOD policy goes further than that,
requiring the evaluation of all IA and IA-enabled products.
While vendors are primarily driven by product cost,
functionality and time to market, security has also become a
significant consideration. Recently the largest vendors have
pledged to make security a priority. The decisions of those
vendors are based on thorough business cases analyses. None can
afford the continued cost of the race against the ``penetrate
and patch'' approach to deal with latent vulnerabilities in
software packages. The economic cost of that approach is
enormous and does not result in a higher level of security.
Sound software engineering practices like those tested in a
NIAP evaluation are an essential element in the elimination of
vulnerabilities and critical to the reduction of postdeployment
patching.
Still there remains the cost of evaluation and time of
evaluation. Both are functions of the complexity of a product,
the level of evaluation, the quality of a vendor's product and
the vendor's preparation for evaluation. Product complexity in
the evaluation level is directly proportional to the amount of
testing required, and the amount of testing is directly
proportional to the time and cost. A quality product may not
require repeat testing. However, products that do get into a
test, fix and test cycle incur additional cost not only for
testing, but also for product modification.
Some vendors, especially small vendors, are concerned about
the cost and time of evaluation regardless of the product's
complexity. During the development of DOD policy, we met with
small businesses individually and in multivendor forums, and,
based on their input, we developed policy that attempts to
remedy some of their concerns.
The evaluation process does what it was designed to do. It
provides standardized evaluation reports that help make--help
us make informed risk management decisions with respect to the
security of our networks and systems. Expectations of evaluated
products should not exceed what the evaluations are designed to
provide. The type of testing that uncovers vulnerability can be
done by the NIAP laboratories and will be done if required. The
depth of evaluation depends on how much time and how much money
we are willing to pay, as well as how much risk we are willing
to accept. Evaluations do not guarantee security. The security
comes from sound systems engineering, the combination of
technologies, operations and people.
The President's recent National Strategy to Secure
Cyberspace requires a comprehensive review of NIAP to examine
its effectiveness and expansion potential. We are conducting
that review in collaboration with the Department of Homeland
Security. DOD is also investigating the issue of software
assurance with respect to all software, not just IA and IA-
enabled products, again working with the Department of Homeland
Security.
The challenges we face are the same challenges found
throughout government and industry, challenges we are
addressing in our IA strategic plan. DOD is making progress
managing the risks successfully across all of our national
security and defense missions. That success is documented in
our FISMA reports as well as our annual IA report to Congress.
Most importantly, however, it's reflected in our ability to act
as an enabler and not as an impediment in the conduct of
networkcentric operations in several theaters across the globe.
I appreciate the opportunity to appear before your
committee and look forward to your continuing support on this
very critical issue. Thank you very much.
[The prepared statement of Mr. Gorrie follows:]
[GRAPHIC] [TIFF OMITTED] T2771.035
[GRAPHIC] [TIFF OMITTED] T2771.036
[GRAPHIC] [TIFF OMITTED] T2771.037
[GRAPHIC] [TIFF OMITTED] T2771.038
[GRAPHIC] [TIFF OMITTED] T2771.039
[GRAPHIC] [TIFF OMITTED] T2771.040
[GRAPHIC] [TIFF OMITTED] T2771.041
[GRAPHIC] [TIFF OMITTED] T2771.042
[GRAPHIC] [TIFF OMITTED] T2771.043
Mr. Putnam. Thank you very much. And we are delighted to
have been joined by the ranking member of the subcommittee, the
gentleman from Missouri Mr. Clay, and the distinguished
gentlelady from California Ms. Watson. And at this time I will
recognize the ranking member for his opening statement.
Mr. Clay. Thank you, Mr. Chairman, and especially for
calling this hearing.
I'd like to reiterate two points that I made at last week's
hearing. First, the government should use its power in the
computer software marketplace to acquire safer software.
Second, software vendors should be more aware of the security
configuration of the software they produce. Let me briefly
elaborate on these two points.
The Federal Government spends billions each year on
computer hardware and software. Those purchases have a strong
influence on what gets produced and sold to the public. The
Federal Government can use its market power to change the
quality of software produced by only buying software that meets
security standards. The result will be an increase in the
security of all software and better protection for the public.
This is a simple formula. The government doesn't have to
regulate software manufacturers, it only has to use its
position in the marketplace.
Mark Forman, the former Federal CIO and regular witness
before this subcommittee, incorporated an idea similar to this
when he developed the Smart-Buy program. Mr. Forman realized
that Federal agencies were buying the same software over and
over again. Each agency was paying a different price for the
same software, and the Federal Government was getting little or
no leverage out of its position in the marketplace. No business
would operate like that.
I believe we should build on Mr. Forman's idea to buy not
cheaper software, but better software. I hope the new CIO,
Karen Evans, will work with the subcommittee to incorporate
this concept into the Smart-Buy program. We don't have to wait
for computer companies to develop new security procedures.
There are some steps that can be taken very quickly to improve
computer security. We saw this earlier this year when Microsoft
began shipping software that was configured differently.
The story Microsoft tells is that the company realized that
it was shipping software with all the gates opened. A good
computer manager systematically went through the software,
closing gate after gate. Those with less training left the
gates open, and the hackers walked in.
Shipping software with secure configurations should be a
first priority of all computer companies.
I look forward to the testimony today of these witnesses,
and I hope that our witnesses will consider my suggestions and
provide the committee with their comments on it.
Thank you, Mr. Chairman, for yielding.
Mr. Putnam. Thank you very much.
[The prepared statement of Hon. Wm. Lacy Clay follows:]
[GRAPHIC] [TIFF OMITTED] T2771.044
[GRAPHIC] [TIFF OMITTED] T2771.045
Mr. Putnam. At this time we will recognize the gentlelady
from California Ms. Watson for her remarks.
Ms. Watson. Thank you so much, Mr. Chairman. I do
appreciate this opportunity.
Over the last decade or so, the Internet has become a force
in our society that it is difficult to identify critical
networks in our Nation that are not connected to the Internet.
Electricity, traffic, water, freight, all these systems rely on
the Internet for their function. This reliance on the Internet
has yielded tremendous gains in efficiency, yet we are
constantly reminded of the vulnerabilities inherent in such
reliance on the Internet.
Most recently, the Blaster and the SoBig viruses posed
major challenges to the integrity of America's infrastructure.
Thankfully, none of the cyber attacks known to us have resulted
in cataclysmic damage to the United States, or to our people,
or to our infrastructure, at least not yet. We have had many
close calls. And in the wake of September 11, many analysts
familiar with global terrorism blame America's leaders for
missing the signs that we were vulnerable to conventional
terrorism. If we in Congress do not wake up to the clear
warning signs of our vulnerability, we would be committing just
as grave a mistake.
In my experience in Micronesia, in my embassy, is that we
were getting warnings by cable from the State Department on a
daily basis of a virus that ran through our most sensitive
computers in the embassy. That's a very scary notion when you
depend on the Internet 24/7 to communicate.
And so this hearing is very, very valuable to the basic
security of our country, and I really would like to be here to
hear every bit of the comments that are being made by the panel
with such expertise. But we have a hearing on terrorism, and I
do hope our enemies around the globe do not--are not able to
master the Internet to the extent that they know more than we
do and they can get our country's secrets.
So thank you so much, panelists, in bringing your expertise
to us. Thank you, Mr. Chairman. And I'm going to run down to
that classified hearing.
Mr. Putnam. You just throw us off like a bad habit.
Ms. Watson. I want to find out what those real secrets are.
Mr. Putnam. We will begin with the questions for the first
panel.
Mr. Gorrie, could you explain why DOD decided to adopt the
Common Criteria requirement for all DOD procurement. What led
to that decision?
Mr. Gorrie. The original NSTISSP 11 requirement was for--
only for national security systems, and if you look at the term
``national security systems,'' that's a legislative construct
that was brought into being in I believe it was the Nunn-Warner
amendment to the Brooks Act. The Brooks Act established that
all ADPE, automatic data processing equipment, would be bought
through GSA. The Department of Defense found that GSA wasn't
really responsive to that. This was in 1986. And the Warner
amendment, as it was known, changed that to reflect the term
``national security systems.'' so it said that--and I have it
here somewhere, but--and I'll get it for you later, but it says
all national security systems--and it went on to list what a
national security system was: anything that handled classified
information, did intelligence, did cryptologic work, did
weapons systems, were national security systems, with the
exception of what they term support systems, which were
personnel systems, logistics systems and things of that nature.
If you went out and talked to any commander around the
globe and asked them if their personnel system or their
logistic system was a critical part of their warfighting
capability, they would undoubtedly all say yes, and so that is
why--the reason we in DOD said you need not only security in
your national security systems, but in all other systems that
we use, because they all touch one another, and the potential
for a security flaw in one could spill over to other ones.
Mr. Putnam. How would you evaluate your experience thus far
in terms of the weaknesses of it, the strengths of it, lessons
we can derive as we contemplate its usage beyond DOD?
Mr. Gorrie. First, the weaknesses of it. A lot of it has to
do with our interaction with our vendors. Some of the vendors
are not--and even some of the people, the users in DOD who have
to follow the rule, are not familiar exactly with what the rule
entails. Some of the criticisms from small businesses that they
can't realize a return on investment are borne out of ignorance
of what the policy says, because the policy does provide them
the opportunity, when making a contract with the government to
sell their particular product, that the only thing that they
need to do is to stipulate in the contract that they will have
the product evaluated, not that they have to evaluate the
product prior to establishing the contract, which gives them
the opportunity, if selected, to realize a return on that
investment because they can include that cost in the cost to
the government.
The number of systems which are being evaluated, although
adequate right now, needs to be much, much higher, and the
types of systems that are being evaluated need to be expanded.
Those are our problems.
Benefits, as I said in my testimony, the ability to know
what a product will do is one of the biggest benefits we can
have. You can get the glossy brochure from a vendor that says
this is the best thing since sliced bread, but until you put it
to the test, you don't know what that product will do. And an
independent evaluation such as that provided by the NIAP is
invaluable not only because you know what it will do, but when
you certify and accredit a particular system to be able to be
connected to our networks, you have to make a risk decision
whether or not that system is safe enough. If you know exactly
what those products are doing, then you can craft other things
around that particular product to circumvent any shortcomings
it may have, things like an operational procedure or some kind
of policy control or other things. So in that particular sense,
the reports that we get out of NIAP are invaluable in order to
make our systems safe.
Mr. Putnam. Thank you.
Mr. Fleming, when you developed NSTISSP 11, the requirement
that national security systems purchase software certified
after the Common Criteria, what consideration was given to its
impact on small business?
Mr. Fleming. NSTISSP 11, first of all, comes out of the
national security community, which comprises of some 21 or 22
Federal departments and agencies and sort of the national
security slice across those agencies, where in DOD it might go
deeper than it would in some of the other agencies. It also
requires an evaluation of all information assurance products.
The NIAP process is only one of the schemers. NSA does
evaluations for high-grade cryptography. So the NSTISSP 11
applies to a broader thing than just NIAP.
As far as small businesses are concerned, the cost of
evaluation, as I mentioned in the testimony, varies
considerably depending on the assurance level. And when NSTISSP
11 was originally issued, it did not specify that all products
had to be evaluated in the beginning. It put a date in there of
July 2002. It came out in 1999. There was a period in there
where it was something to be considered. And the idea there was
to allow companies to get used to both the process and the
profiles that were coming out. So the mandate did not start
until, in fact, almost 2 years later in 2002. So the idea was
to allow companies to grow toward what this was.
The second thing was it didn't specify any particular
evaluation level. The beginning thinking was any evaluation
level is better than none. And so the cost is, in fact,
considerably lower at the lower levels than it would be at the
higher levels because of the demand for generating evidence. So
the idea was to ramp this process up to allow companies to grow
with it and, over time, ever increase the assurance level in
these products.
So that's how we wanted to consider, in fact, all vendors,
but in particular the small companies.
Mr. Putnam. In the beginning of your answer, you mentioned
that this was to cut across national security systems. Does the
Justice Department and Homeland Security and State also utilize
government criteria?
Mr. Fleming. Yes. NSTISSP 11 includes all those agencies
and the opportunity for them--obviously NSTISSP 11 applies at
that level the opportunity to use the Common Criteria, and the
NIAP process is there for any buyer. But, yes, NSTISSP 11
covers those kinds of agencies for their national security
systems.
Mr. Putnam. Mr. Roback, as all of us are aware, and we have
held hearings related to this, the Blaster worm exploited a
flaw in Microsoft's operating systems to infect thousands of
computers. Since that system was certified, why wasn't the flaw
found? What is the weakness in the evaluation that does not get
at code flaws?
Mr. Roback. I think you have to look at the range of
possibilities that the NIAP testing program offers. At the low
end, where you are looking at things like documentation of how
the product was developed, you are not getting into the very
detailed code review that you get at the very, very high levels
of assurance. So it sort of depends on what level you want to
pick for your evaluation, which is the flexibility of it. A
vendor can bring in product and target any one of the seven
levels or create their own. So unless they target something at
the very high level, which, by the way, costs a lot more and
takes a lot longer, you are not going to get that level of
review. And even if you do, it's subject to human subjectivity
in the review.
So you may not get--because we don't have very specific
standards for this, and you probably couldn't at that level for
millions of lines of code--standards you can do very quick,
very exact testing. So there's' some art in here, too.
Mr. Putnam. Thank you very much.
Mr. Clay.
Mr. Clay. To all of the witnesses, I would just like to
hear from you or hear your comments on the proposal to add
secure configurations as another dimension of a Common
Criteria. Is this feasible, and how long would it take? We'll
start with you, Mr. Roback.
Mr. Roback. Actually under the Cybersecurity R&D Act that
was passed by the Congress late last fall, they assigned to
NIST the task of developing security configurations for
specific IT products. And so we are holding a workshop later in
September trying to invite the vendors in, and other Federal
agencies and NSA and others have already developed some of
these checklists for some very specific products. So some of
these do exist.
Actually I think it would be a very good thing, because if
you look at a spectrum, first you want to have very strong
standards. Then you want to have some testing program that
tells you whether the standard was correctly implemented. And
third, you want to have those configurations so that when a
system gets one of those products, they know where to set the
settings, because even if they are shipped from the vendor with
security turned on, which is not always the case, but sometimes
it is, it is not necessarily always right for the environment
that it's being put into.
Configuration guidance is a very good thing. It's also
important to remember there's a range of potential
environments; that is, the security you would have for a home
user might be very different from the security at NSA or the
security of a large, centrally managed enterprise. So you have
to keep that in mind, too, there's a range there, because
there's a range of risks in the type of information that's
being exposed.
So it does get complicated, but I think checklists of that
sort are very useful.
Mr. Fleming. I would agree with everything Mr. Roback said.
This is a life cycle. Security is a life cycle endeavor. It
just doesn't stop when the product is certified and goes out
into the field security. Every day you've got to watch these
products, particularly security products that sit sometimes in
the way of system performance, and it is so often tempting to
tweak that firewall a little bit to allow the bandwidth to get
greater, but you may have, in fact, left open the door you
don't want to open.
So I would add to Mr. Roback's points the human dimensions
of this. It boils down to how well trained is that system
administrator or that security administrator; how well do they
understand the multitude of configurations that these products
can, in fact, take, and which ones are the good ones and the
bad ones. So there's a dimension in this of awareness and
training of individuals along with the ideas that Mr. Roback
put forth in terms of having configuration guides. And we have
been a very, very strong partner in the generation of these
configuration guides for major IT systems, but there are many
other technologies that need a similar kind of guide for a
well-meaning, but sometimes difficult job called system
administration, security administration.
Mr. Clay. Mr. Gorrie, anything to add?
Mr. Gorrie. Well, I could attest that it does work. In DOD
we have been using secure technical implementation guides
[STIGs], for our products for our operating systems and other
things for a long time. We have a process that is known as gold
disk, where we will go out and put particular security settings
on operating systems.
STIGs, security technical implementation guidance. We have
them in DOD. They work, and it depends on, again, as Mr. Roback
stated, what environment you want to use them in. If you are
using them for an inventory system in a gym, no sense in
tightening it all down because you want to be more open and
share those sorts of things. If you have a critical system that
you need protected, it needs to be ratcheted down. And all the
people who participate in that network have to have it
ratcheted down to the same degree.
Mr. Clay. In your opinion, would it be possible to certify
software configurations separate from the Common Criteria
evaluation?
Mr. Gorrie. I don't know. I would have to defer on that.
Mr. Clay. Anybody?
Mr. Roback. Well, I am not sure if certification is the
precise word, but I think that there are indeed--you can
separate the two. Whether a product has been tested to know
whether the security features work correctly is a separate
question from where to turn on and turn off the security
features.
However, if you haven't gone through certification, the
testing process, you are not going to have a great deal of
assurance that even if you turn something on, the security is
working. And the example I like to give is you go to a Web
site, and you get the little lock in the corner on your
browser. Well, why do you have any confidence whatsoever that
it is doing anything other than showing you a little picture of
a lock? If it hasn't gone through testing, you really don't
know, other than it makes a nice little picture in the corner.
So that is why testing is so important in addition to
turning on the security.
Mr. Clay. For all of the witnesses, again, I would like
each of you to comment on my proposal that the Federal
Government use its market power to improve the level of
security for our purchasers. Do you believe this is feasible?
Mr. Fleming. I will start. During the testimony I used a
phrase called converged market, and I think it is along the
lines of your reference back to the Smart Buy Program that Mr.
Foreman put in place.
The idea of a converged market would be find that level of
security goodness, that assurance level and that set of
security mechanisms that a large buying sector could agree to,
the DOD, the national security community, the other Federal
agencies, the critical infrastructure marketplace that Ms.
Watson referred to, such that a vendor would see a return on
investment good enough for them to shoot for that level.
And so this idea of getting a common level that all would
buy into would, I think, be a good incentive for vendors. Make
it appropriate so it is not a bridge too far, and then
standardize on that level and let vendors shoot for that level
so you can get this economy of scale.
Mr. Clay. Thank you.
Mr. Gorrie, one question from our other Member who had to
leave. She says: It is good to hear that you understand the
business costs of the reactive plug and patch approach, but how
widespread do you feel this view has been accepted throughout
the technology industry? What can we do to spread this message
and change the approach?
Mr. Gorrie. I think if you will ask the panel members that
follow us, those are the words that were given to me by them. I
mean, they were the ones who told me those things. I didn't
make that up.
How can we spread it? I think it will spread itself.
Vendors whose products are well developed and have fewer
problems as far as having to go out and patch them and things
of that nature will be bought more. People will see the benefit
of buying them, not being able to be hacked, not having to go
in and reengineer their systems every time a patch comes out.
Those who have products which are constantly being patched
will find their position in the marketplace becoming lower. It
is a self-regulating system, and it will become more so in the
future as more and more patches have to be made to accommodate
shortcomings in software.
Mr. Clay. OK. I thank the panel.
Thank you, Mr. Chairman.
Mr. Putnam. Thank you, Mr. Clay.
Mr. Fleming, under Common Criteria evaluation, the product
is tested by itself. Obviously it will be used in conjunction
with a variety of other products. Is that taken into
consideration at all? And how is that issue resolved in terms
of the impacts or the problems that can occur with the
connectivity?
Mr. Fleming. Good. First of all, the product is typically
tested in an environment. In order to make the test meaningful,
there is an assumed environment. However, that may not
represent all possible environments for the application in the
real world.
So there is--and these are my terms. There is this ``little
c'' certification, which is certifying that the product is
doing what its claim is. Then there is the application of that
product, along with other products, into a larger system. So
there needs to be another certification, which is at what I
will call the ``big C,'' at the system level. There are
processes in place in the DOD and across the Federal
communities that go by somewhat different names, and they are
like kind of complicated names like DITSCAP, Defense
Information Security Accreditation Program, but that is a
system-level look. So I see the Common Criteria and any other
evaluation program at the product level generating evidence
about the performance of the various components. Then there
needs to be a separate look at the much larger level, for what
the total system security certification is.
Now, I will state that the calculus for that is somewhat
difficult, because it is not just saying product A has this
level of goodness, product B has this level of goodness. You
just don't add A and B and get C. In fact, it is a much more
complex relationship when you start bringing many products
together. But, nevertheless, there are processes in place in
the DOD and beyond the DOD to bring this larger certification
into play toward the ultimate accreditation of that system to
operate in a real environment.
Mr. Roback. If I can just add to that, that question of
when you understand the property of one component and then the
property of another, and you put them together, that is what
the researchers call the composability issue, trying to
understand in a rigorous way what you have when you put them
together and add them up.
If I could just add to Mr. Gorrie's comment earlier about
the software quality and patching, I think one of the problems
we face is the whole developmental cycle in the industry of
software products. And you have to really look at how software
products are developed in a rigorous sense of specifications
and so forth.
If you really want to improve the overall security and get
away from this problem of continually chasing our own tail and
trying to patch, this is a Web site where we put up
vulnerabilities in commercial products so people can learn
about them and go find fixes for them. Right now we have over
6,000 vulnerabilities up there.
And, you know, the tolerance of the marketplace for these
products that come out with flaws is just astounding. We really
need to look at the overall quality issue of the products; not
just the security, but the overall quality. Do they do what
they are intended to do? It is a challenge.
Mr. Putnam. Mr. Gorrie, several testimonies mentioned a
waiver process for the Common Criteria. Under what
circumstances would a waiver be granted?
Mr. Gorrie. With the way that we have constructed policy
within the DOD, I would find it very rare that you would have a
request for a waiver. The only one that I could think of would
be in a situation if we were going to war, we needed a
particular product because its security features were just so
obviously great that we could not not afford to have it in our
system. But even then, the way that we have policy built, it
says that you need not to contract to purchase that piece of
equipment, you need not have it evaluated, only have it in the
contract that you will have it evaluated as a condition of
purchase.
However, the vendor always has the option to say, you may
need this, this may be the best thing since sliced bread, but
we are not going to have it evaluated. If we needed that
product that bad, then the user of that product, or the person
who wanted to put that product into their system, would have to
petition for a waiver, and then we would either have it
evaluated internally within DOD through some process, or just
use it because it was so important to use. But in any regular
process I--because of the way policy is written, I do not see
the need for waivers.
Mr. Putnam. My final question for this panel will be this,
and we will begin with Mr. Roback: Should the Common Criteria
certification be extended to cover the entire Federal
Government?
Mr. Roback. That is a good question, one we are often
asked. Let me just start by mentioning that it is policy for
the nonnational security side of government that cryptographic
products have to go through testing, and there is no waivers
allowed for that under FISMA.
I don't think the question is necessarily should we adopt
that policy, yes or no. There is actually quite a range of
options between doing nothing and adopting that policy, and
even things beyond that policy. So you might ask yourself:
Well, maybe it doesn't make sense to say something can be
certified against any specification that is brought forward,
but maybe what we need to do is look at things like once we
have good specifications for specific technology, that if an
agency is buying that technology, they should be buying
something that has been evaluated against those specs.
So not just that you can bring in--I think someone in their
testimony talked about a product that paints the screen blue,
and it can go through and get a certification. Well, I don't
know if those products are going to do us any good. So I think
there is some range of options we have here, and we really need
to look at those. Rather than just say, that is the policy for
national security; we should simply adopt it.
I think we need to learn more from the experience as well.
Is it really pushing the vendors toward more security or not?
Mr. Putnam. Mr. Fleming.
Mr. Fleming. We are putting our trust in networks, in
things called security products. They have become sort of a
foundation piece, a trust anchor, if you like. And so it would
seem to me we should take extraordinary measures, not
necessarily expensive, but take extra measures to ensure that,
in fact, that trust is well founded.
So having some rigor in how we look at security products
is, I think, important. Independent evaluation is an important
piece of that rigor. That is something different than the
vendor claims. So where does one get this independent look?
What is the most efficient way to get that independent look
that in and of itself can be trusted by people who use these
systems?
So whether it is a Common Criteria-based system that we
use, whether it is some derivative that may be the result of an
evolution of the process, I believe that we need to put honest
faith in our security products through some independent
specification evaluation process. It is too important just to
sort of leave to the normal process.
Mr. Putnam. Mr. Gorrie.
Mr. Gorrie. There is two parts to this, as Mr. Fleming
said. There is the independent evaluation portion of it, the
Underwriters Laboratory, if you will. Should that be extended
to the rest of the Federal Government? The Department of
Homeland Security thinks it is. That is why they want us with
them to do a review of the NIAP process, to see what that
extensibility of the process is to the rest of the Federal
Government.
Is it extensible to the rest of the civil population? No
one forces a consumer to buy a lamp that has the Underwriters
Laboratory stamp on the cord, and perhaps no one should be
forced to buy a piece of IT security equipment with a NIAP
certificate associated with it.
There is the evaluation program itself, and then there is
the regulatory and policy piece that goes along with it. And
although I think the evaluation portion of it can go forward,
because knowledge is power, the--how you instantiate that in
regulation and in national policy is a different matter
altogether.
Mr. Putnam. Thank you very much. And I want to thank all of
our witnesses on this first panel and encourage you to stay and
listen to the second panel, if your time and schedule allow.
With that, the committee will stand in recess for 2 minutes
while the first panel is dismissed and the second panel is
seated.
[Recess.]
Mr. Putnam. The committee will reconvene. Before we swear
in the second panel, I did want to announce publicly that the
executive session on SCADA, which was scheduled for tomorrow by
the subcommittee, has been postponed thanks to Hurricane
Isabel.
And with that, I would ask panel two to please rise and
raise your right hands for swearing in.
[Witnesses sworn.]
Mr. Putnam. Note for the record that all of the witnesses
responded in the affirmative, and we will move immediately to
their testimony. Again, I would ask that you limit your remarks
to 5 minutes, and your entire written statement will be
submitted for the record.
Our first witness is David Thompson. Mr. Thompson directs
the CygnaCom Security Evaluation Laboratory. He has led a team
to support certification for the Air Force Scope Command's High
Frequency voice and data communications system, and managed
Public Key Infrastructure products at several Department of
Energy National Labs. He led a team to write a Common Criteria
security target for Red Hat Linux 5.1, and helped translate
high-assurance criteria into Common Criteria protection
profiles.
Previously Mr. Thompson evaluated the security of network
and computing configurations for the space station and space
shuttle, and assessed proposed uses of cryptography and
distributed authentication at NASA. He was session chairman for
the 1993 AIS Security Technology for Space Operations
conference, and served on a board investigating a software
configuration management failure in a space shuttle mission.
Welcome to the subcommittee. You are recognized, Mr.
Thompson.
STATEMENT OF J. DAVID THOMPSON, DIRECTOR, SECURITY EVALUATION
LABORATORY, CYGNACOM SOLUTIONS
Mr. Thompson. Thank you. I would like to thank the
committee Chair and all of its members for their interest in
this issue and their leadership.
The motivation for product testing that led to the creation
of the Common Criteria came from the U.S. Government's
certification and accreditation process for systems. Most
systems included at least one computer with operating systems
that needed a security functionality identified and assessed.
Since operating systems are quite complex and have many key
security functions, considerable effort is required to do an
appropriate security assessment.
As computers became more commodities, the notion of
performing these difficult evaluations once and using the
results in repeated CNAs took hold. In the early 1990's, as the
expense of having products evaluated to different security
criteria in different countries increased, Western governments
began to seek a set of Common Criteria that they could endorse.
We are still in the early stages of implementing the resulting
Common Criteria. But the original government participants are
still actively engaged, and additional governments are getting
involved.
Industry also begun to see the value of a common security
performance process. The CC defined seven sets of security
assurances called evaluation assurance levels. EAL1 has the
least assurance, and the EAL7 the most. The most commonly used
assurance levels are EAL2 and EAL4. The EAL2 is an acceptable
assurance level for most products, and EAL4 is often specified
for products that are employed in the first line of defense,
such as firewalls and operating systems.
Custom sets of CC assurances can also be chosen when one of
the seven EALs is not precisely suitable. The result of a
successful CC evaluation is a published security target that
precisely documents the security functions that the product
claims to meet and establishes in precise terms whether these
claims are true, the security target to be used to determine
the product's suitability for a particular use and to compare
its security functionality with that of other products.
It is practically impossible to determine that a product of
any complexity will be secure regardless of its configuration
or that security will mean the same thing in all the situations
in which the product is used. What CC testing does show is that
to the specified level of assurance, the security functions the
vendor claims the product has work as described, and that a
coherent and mutually supported set of security functions is
available.
Just because a product has been successfully evaluated
under the Common Criteria does not mean that it has no
vulnerabilities. Instead, it shows that the product is suitable
for use as a component of a secure system. It is primarily
focused on design and development process issues.
Although higher CC assurances, such as the EAL7, also can
significantly reduce the possibility of bugs, the Common
Criteria evaluation process has several strengths. It provides
consumers with an independent and well-monitored assessment of
vendor security claims. It provides a precise description of a
product's security features that is readily comparable to those
of other evaluated products. It assesses the ability of a
product to be used to build secure systems. It demonstrates
that at least one configuration of a product meets the claimed
security requirements. It allows precise tailoring of the
security criteria to the capabilities of products. It uncovers
design flaws and sometimes software bugs. It focuses vendors on
security issues. It constitutes the most rigorous and thorough
independent product testing process commercially available. It
provides international mutual recognition so that vendors have
to pursue only one evaluation against a single criteria.
The Common Criteria evaluation process also has some
drawbacks. It creates additional expense for product vendors.
CC evaluation is applied to an exact version of a product in a
precise hardware environment, making it sometimes hard to field
a product that is strictly conformant. As consumer protection
profiles evolve, as vendor products are revised, they must be
reevaluated. The evaluation process is complex and time-
consuming, which means it requires a lot of vendor resources,
understanding and participation. Some of these are conflicting.
For example, establishing the security of a product across a
broad range of its configurations among many versions is more
difficult and would further increase the expense.
While large vendors are more easily able to absorb the cost
of an evaluation than smaller vendors, small vendors benefit
more from an independent product assessment that makes it
easier for customers to compare its products' security features
to those of its better-known competitors.
The CCE offers a broad range of assurances and a
corresponding broad range of costs. The EAL1 evaluation costs
are in the low tens of thousands of dollars. EAL1 is adequate
for many applications. Higher assurance has higher cost and is
appropriate where the security risks are higher. The problem of
eliminating security bugs from complex systems, such as those
we read about regularly in the news, requires resources many
orders of magnitude greater than those required for CC
evaluations. There is little theory to support solutions to
this problem, and it remains an art form.
The most productive approaches to bug elimination involve
improved software engineering practices to prevent the
introduction of bugs in the first place. Finding and fixing
bugs in existing products is always more expensive.
The CC product evaluation process is a very effective tool
when used in the right context. The international support and
precise specification of security attributes minimizes the
problems inherent in integrating diverse systems and components
built in different countries and secure systems whose security
attributes are well understood.
The Common Criteria evaluation, however, does not serve
every purpose. The fact that attempts are made to apply it to
situations for which it was not designed shows how great is the
need for other kinds of security testing and the challenges
facing the available security evaluation services.
Thank you for this opportunity to address you today.
Mr. Putnam. Thank you very much, Mr. Thompson. We are glad
to have you.
[The prepared statement of Mr. Thompson follows:]
[GRAPHIC] [TIFF OMITTED] T2771.046
[GRAPHIC] [TIFF OMITTED] T2771.047
[GRAPHIC] [TIFF OMITTED] T2771.048
[GRAPHIC] [TIFF OMITTED] T2771.049
Mr. Putnam. Our next witness is Mary Anne Davidson. Ms.
Davidson is the chief security officer at Oracle Corp., where
she has been for the last 14 years. As Oracle's CSO, she is
responsible for Oracle product security, corporate
infrastructure security and security policies, as well as
security evaluations, assessments and incident handling.
Ms. Davidson also represents Oracle on the Board of
Directors of the Information Technology Information Security
Analysis Center, and is on the editorial review board of the
Secure Business Quarterly.
Prior to joining Oracle in 1988, Ms. Davidson served as a
commissioned officer in U.S. Navy Civil Engineer Corps, during
which time she was awarded the Navy Achievement Medal. She has
a BSME from the University of Virginia, and an MBA from Wharton
at the University of Pennsylvania.
We always appreciate your interaction with this
subcommittee and your direct and candid remarks. Welcome. You
are recognized.
STATEMENT OF MARY ANN DAVIDSON, CHIEF SECURITY OFFICER, SERVER
TECHNOLOGY PLATFORMS, ORACLE
Ms. Davidson. Mr. Chairman, Ranking Member Clay, on behalf
of Oracle, I appreciate the opportunity to be here today to
offer Oracle's perspective on the Common Criteria.
Oracle is uniquely qualified to comment on information
assurance policies. We have spent more than 25 years building
information management systems for customers that I
affectionately call the professional paranoids, which include
U.S. intelligence agencies and the Defense Department.
To gain and maintain the business of the most security-
conscious customers on the planet, we have made extraordinary
investment in information assurance and have 17 independent
security evaluations to show for it, with 4 more in process.
The collective impact of Code Red, Blaster and SoBig to our
economy, which amounts to billions of dollars in repairs and
down time, have worked to send a sobering message: It is long
past time for the entire Federal Government to get serious
about information assurance. The benefits go beyond secure
Federal information systems. A strong Federal information
assurance policy has a potential to change the entire software
industry for the better. Let me tell you there is no vendor
when faced with this requirement who will build two versions of
software, one that is strong, robust and well-engineered, and a
buggy, crummy version for the commercial sector.
Fortunately, some Federal agencies are listening. NSTISSP
11 and DOD Directive 8500.1 draw a constructive, clear
prosecurity line in the sand. The question is not whether
NSTISSP 11 makes sense or not. We have had that debate, and it
is over.
The NSTISSP 11, with the linkage to the Common Criteria,
the de facto worldwide evaluation standard, is making a
positive, constructive difference in software development. The
Common Criteria has three key benefits for vendors who do
evaluations: You have more secure products. Evaluators find
security vulnerabilities which must be remedied prior to
receive a certificate. There has been a lot of discussion about
the cost of evaluations, but I have done the analysis, and if
we find or prevent even one significant security flaw in our
products going through the evaluation, it more than pays for
the cost of the evaluation, even at the highest assurance
levels that are viable for commercial software, which says
nothing about the expense to customers that we prevent by
getting it right the first time.
Second, a more secure development process. Evaluations
actually force you to have a secure development process
throughout the entire development process. Security can't be
something that is thrown on in the last 2 weeks of a cycle and
has to be baked into the development process. That is what the
evaluators are looking at.
Third, and probably most important, a strong culture of
security. If you do evaluations as part of development, then
security becomes baked into your corporate culture. That is the
biggest problem that we have in the industry: Security always
seems to be someone else's job. At Oracle I tell our
developers, you are personally responsible and accountable for
every single line of code you write.
Since NSTISSP 11 has gone into effect, we have seen very
positive developments. More firms are doing evaluations. Firms,
including Oracle, are sponsoring open-source evaluations. Many
other industries are looking at certification efforts along the
same lines as the Common Criteria. This has been successful
because industry believes the Federal Government is serious
this time, and that is a major victory. And thanks goes to
people within DOD, the Intel Community, and Congress, who are
making an effort to make the process work.
So what can we do to make this process work better? You can
hold the line by maintaining an eternal but pragmatic vigilance
through the no-waivers policy. I said a year ago that it was
time for the government to chirp or get off the twig on
information assurance, and there has been a lot of chirping
going on.
But there are still those who want to get off the twig by
getting a waiver or seeking opt-outs. It sends a bad message to
the marketplace to say that NSTISSP 11 does not apply to us. It
really needs to apply to intelligence across the board.
NSTISSP 11 should be extended beyond traditional national
security systems, and specifically, I think DHS should look at
applying this to their own systems. Clearly they have a mission
of national security.
NSTISSP 11 shouldn't be allowed to--protection profiles
should not be agency-specific wish lists. I think vendors are
willing to do an evaluation against a common protection
profile, but they are not willing to do three of them per each
special agency.
Country independence of laboratories should be maintained.
We do our evaluations in the United Kingdom, because the cost
is lower and the expertise is actually far higher than we have
found in the United States for our particular product set. We
still get resistance to foreign evaluations, and this is
ridiculous. We are very happy to support U.S. labs as a
competitive alternative, but competence knows no national
boundaries.
A couple of final points. There are three things that the
government can do to foster better security beyond evaluations.
We know that it does not provide a silver bullet or perfect
products. The Federal Government should require that products
have a default setting that is secure out of the box. I think
NIST can do a lot of work here. This would also provide a lot
of immunity to a number of viruses and worms, because more
systems would be locked down by default. It would lower the
cost of operations for the government and other customers.
The government should invest in cybersecurity research.
Quite honestly, the reason vendors cannot find more faults in
the products in development is because the tools do not exist
to do so, and the venture capital community will not fund it
because there is no way to make money on it. If we can stomp
out smallpox through investing in medical research, we can
certainly get rid of buffer overflows. It is just code.
Finally, industry can do more to improve the security
profession. I fully support an alternative to Common Criteria
evaluations, for example, for consumer products where it is
perhaps inappropriate to do a Common Criteria evaluation. And
an example would be the Underwriters Lab. Most products are
just designed to be secure. And Cuisinarts are designed, for
example, that you can't lose your fingers by sticking them in
while the blades are whirring. They are just secure. Consumers
don't have to do something special to make them operate
securely.
NSTISSP 11, DOD 8500.1, and the national strategy are
welcome developments because they are moving the debate to the
expectation that everything will be secure. I believe we have
turned a corner, but it took 10 years and numerous sobering
events to get us here. It will take continued vigilance and
continued leadership here in Congress to keep us on this road.
Thank you again, Mr. Chairman, for the opportunity to
testify today.
Mr. Putnam. Thank you very much, Ms. Davidson.
[The prepared statement of Ms. Davidson follows:]
[GRAPHIC] [TIFF OMITTED] T2771.050
[GRAPHIC] [TIFF OMITTED] T2771.051
[GRAPHIC] [TIFF OMITTED] T2771.052
[GRAPHIC] [TIFF OMITTED] T2771.053
[GRAPHIC] [TIFF OMITTED] T2771.054
[GRAPHIC] [TIFF OMITTED] T2771.055
[GRAPHIC] [TIFF OMITTED] T2771.056
Mr. Putnam. Mr. Clay, if I heard her correctly, she said
that she tells her developers that they are personally
responsible for every line of code that they write. It is a
good thing nobody holds us to that standard on the U.S. Code.
Our next witness is Mr. Klaus. Christopher W. Klaus is the
founder and chief technology officer of Internet Security
Systems, Inc., a leading global provider of information
protection solutions that secure IT infrastructure and defend
key online assets from attack and misuse. Prior to founding
Internet Security Systems, Mr. Klaus developed the Internet
Scanner, the first vulnerability scanner, while attending the
Georgia Institute of Technology.
Mr. Klaus was honored in MIT's magazine, Technology Review.
In addition, he received the award for Ernst & Young's
Entrepreneur of the Year in 1999, in the category of Internet
products and services.
Welcome to the subcommittee. You are recognized.
STATEMENT OF CHRISTOPHER W. KLAUS, CHIEF TECHNOLOGY OFFICER,
INTERNET SECURITY SYSTEMS, INC.
Mr. Klaus. Thank you, Mr. Chairman. It is an honor to
testify today. And I am representing Internet Security Systems
from a small company's point of view that builds the security
products, and we are in the process of going through the Common
Criteria and NIAP certification and wanted to share some of our
experiences as a company going through it, and what are some of
the benefits and some of the failures of the process that we
see today.
We do believe the overall goal and the intent of the Common
Criteria and going through NIAP certification is a positive
goal, but we see that there is at least three areas of major
improvement that need to happen. And if they do not get
addressed, we believe that following this path of requiring the
government to follow the guidelines of NIAP certification
actually makes the government less secure. And we will go
through these three reasons and talk to why do they make both
the government and others that follow the certification less
secure.
No. 1 would be the accuracy. The current different levels
of evaluation do not reflect whether the security product is
actually more accurate in protecting against vulnerabilities
and exposures. To take a step back, let me explain two goals
within security, so you understand what we are measuring.
There is two major goals in security. One is to allow good
people into the network, or into an operating system, into an
application. And what we typically think of good guys in
technology is like your user name and password that allows you
to get into the system. There is biometrics, fingerprinting,
VPN, virtual private networks. All of those technologies are
great for--help certify the right people get into the system.
And one of the problems, though, is it assumes that the
infrastructure stops the bad guys out. So the second goal of
security is keeping bad guys out. The problem we find is that
the assumption that the infrastructure keeps the bad guys out
is false. We know there is everyday bugs in the code. These
bugs lead to vulnerabilities that then allow intruders, worms,
viruses to leverage that.
So on the second goal of keeping the bad guys out, that is
a major area of measurement. And one of the things that we
track very closely as a company that produces security
products, we are tracking over 200 plus vulnerabilities every 3
months, every quarter, as we measure that, and what is
interesting about this is as we go through this process,
products that are less accurate in finding those
vulnerabilities have the same certification as the companies
that have much more accurate products. And if you likened it to
antivirus, which most people are familiar with antivirus
software, if only 10 percent of the vulnerabilities were--or 10
percent of the viruses were found with one product, and 99
percent were found with another product, today they would be
measured equal in terms of the certification level. And that is
one of the major reasons why government agencies that believe
they are getting a more robust product may end up--just because
they are purchasing a higher level certified product may
actually end up with a less robust and less accurate product.
The next major area is speed. The current evaluation
process is extremely slow and bureaucratic. It can take over a
year to become certified. By the time it does become certified,
it is outdated and behind the latest version of protection. The
commercial sector could apply the latest version, and while the
government would lag behind in security, in the race against
cybercrimes threats, all organizations need to apply the
current, most up-to-date security protection products.
I just would add that there is over 40 IDs or intrusion
detection companies in the process today, but only two of them
have actually been certified. So we have a long way to go
before all of them have gone through this process.
The issue, though, also with security products, we are in a
much different stage in the technology industry where we are
rapidly evolving the technology to keep up with the
cyberthreats. A lot of older technologies, like operating
systems, data bases, Web servers, those technologies have been
around longer, more mature, have a much longer development
cycle. So in many cases the larger the application, and the
larger the deployment cycle, the more likely you can keep in
pace with the certification. In the security circles it is at a
much faster rate.
And finally, the cost part of this is that the current
evaluation process is extremely burdensome and costly for
security vendors to follow. And after following the process,
the expense does not--for us has not resulted in any security
improvement. It has not found any buffer overflows. It has not
found anything that many of the hackers and worms and viruses
take advantage of. So, therefore, many of the resources and
capital that we are spending on this, if it was doing something
to make our products a lot better, and more protected, and more
robust, and more accurate, we would be in favor of this.
So those three things, accuracy, speed and cost, are
critical to improving, to make this thing worthwhile.
Mr. Putnam. Thank you very much.
[The prepared statement of Mr. Klaus follows:]
[GRAPHIC] [TIFF OMITTED] T2771.057
[GRAPHIC] [TIFF OMITTED] T2771.058
[GRAPHIC] [TIFF OMITTED] T2771.059
Mr. Putnam. Our next witness, our last witness on this
panel, is Eugene Spafford. Dr. Spafford currently serves as the
director of the Center for Education and Research in
Information Assurance and Security at Purdue University, a
position he has held for 5 years.
He has written and spoken extensively on the topic of
information security. His research focuses on the prevention,
detection, and remediation of information security failures and
misuse, including fault tolerance, software testing and
debugging, and security policies.
He holds a Ph.D. in information computer science from the
Georgia Institute of Technology.
We are delighted to have this level of expertise on the
panel. And you are recognized for 5 minutes.
STATEMENT OF EUGENE H. SPAFFORD, PROFESSOR AND DIRECTOR, CENTER
FOR EDUCATION AND RESEARCH IN INFORMATION ASSURANCE AND
SECURITY, PURDUE UNIVERSITY
Mr. Spafford. Thank you, Mr. Chairman. And thank you also,
Ranking Member Clay and members of the committee.
The question posed to us for this hearing was can the
Common Criteria ensure security for the Federal Government? And
my answer to that is very definitely not. It will not.
And that is not to say that the Common Criteria is not a
valuable instrument. The many thousands of man-years of effort
by experts around the world putting it together has resulted in
a procedure and set of documents that have great value as
guidance for those building systems and for a means to compare
systems as to their level of quality. However, it does not
actually address the problem of ensuring that the government
systems or any systems that possess the certification are
themselves secure. It is in some sense, if I may use the
analogy, similar to wanting to be sure that your house will not
burn down and believing that the Underwriters Laboratory seal
on the cord of your toaster will ensure that. It is not the
case. What it does do is it gives you a small added measure
that one item involved is less likely to cause you damage, but
it is certainly no guarantee for the whole enterprise.
We can see that with an example that has been cited by many
others. If we look at the Windows 2000 operating system, it is
certified at the highest currently available level available
under the Common Criteria, and yet it was a target. It was
vulnerable to the Blaster worm, the Natchi Worm, dozens if not
hundreds of current viruses, and has had nearly 100 patches
issued for it--those are security patches, not functionality
patches--since it was released. And that is something that is
certified at the highest level.
There are other examples. I have given a detailed list in
my written testimony as to why I do not believe that the Common
Criteria is going to give us the level of security that we
want. And in the very limited time available here, what I am
going to do is give a different approach to this, and I am
going to do it by analogy.
Let's take that toaster example that I was talking about.
Suppose that you were the vendor of that toaster, and you
wanted to compete on the market and decided that an evaluation
was something that would give you a competitive advantage. So
you submit it to a consumer testing lab. However, when you
submit it to the testing lab, you submit it without a cord, and
you tell the testing agency that you want to submit it as a
bread storage device.
Well, the agency is required to test it against the
requirements that you gave them. So they will test it as a
bread storage device. They will go against the checklist for
all the devices in the kitchen, and they will discover there
are no radioactive materials or explosives embedded in it, and
that, in fact, it does meet all of the documentation that you
provided, it was built by the engineers in the appropriate way,
and it does store slices of bread. So they give you their
highest level of certification.
You then turn around and put it in the marketplace with a
cord and with a consumer option to include a speaker phone,
because while you are making toast, you want to call your
neighbors and tell them to come over and have some of the
toast. The problem there is that about every tenth piece of
toast that you put in burns and possibly starts a fire. What is
more, the speaker phone is defective and calls up your
neighbors and starts fires in their toasters. And because parts
of the toaster were built overseas in a country that was
unfriendly to the United States, if you use the toaster in any
government kitchen, it simply doesn't work. On top of that, the
manual is badly written. Customers who buy it don't really
understand how to use it. They attempt to toast jello, they use
it in the bathtub. And when all of the various disasters occur,
the fires and deaths, they find that the disclaimer that
shipped with the toaster says that the vendor has no
responsibility for anything that the user may do with the
toaster, and therefore they have no legal recourse, and there
is no penalty that comes back on the vendor.
However, the toaster does very well in the marketplace
because it is cheaper than the other toasters that aren't
certified and happen to work without fault. Those who go out
and buy in large quantity, using the lowest bid, proceed to
make you the market leader.
Certification does not guarantee that what you have is
safe. It says that it meets the standards for the
certification. It also does not tell you that it is going to be
used safely or in an environment where it is appropriate to use
it. That is why the Common Criteria is not appropriate.
Quality needs to be built in from the very beginning. As
has already been noted by others, we don't know how to do that
well, because this is an area that has been underfunded. It is
an area where we need more research. We need more resources put
into the agencies that are involved in this, particularly NIST.
This is a problem that we are going to continue to face for
many years because we have such a large base of legacy code
that is already in place and is too expensive to replace with
something, even if we developed it tomorrow, that was much
better.
Thank you for--the committee for listening to us on this
today, and I stand ready to answer questions.
[The prepared statement of Mr. Spafford follows:]
[GRAPHIC] [TIFF OMITTED] T2771.060
[GRAPHIC] [TIFF OMITTED] T2771.061
[GRAPHIC] [TIFF OMITTED] T2771.062
[GRAPHIC] [TIFF OMITTED] T2771.063
[GRAPHIC] [TIFF OMITTED] T2771.064
[GRAPHIC] [TIFF OMITTED] T2771.065
[GRAPHIC] [TIFF OMITTED] T2771.066
[GRAPHIC] [TIFF OMITTED] T2771.067
[GRAPHIC] [TIFF OMITTED] T2771.068
[GRAPHIC] [TIFF OMITTED] T2771.069
[GRAPHIC] [TIFF OMITTED] T2771.070
[GRAPHIC] [TIFF OMITTED] T2771.071
[GRAPHIC] [TIFF OMITTED] T2771.072
Mr. Putnam. Sounds like this could be a fun panel. Mr. Clay
will lead off this round of questions.
Mr. Clay. Thank you, Mr. Chairman. And I will start with
Mr. Spafford. You said your research center has a data base of
computer vulnerabilities. In your search of that data base, how
many of the software products identified were certified under
the Common Criteria?
Mr. Spafford. Sir, I didn't do a search specifically on
that criteria, but from the numbers that I know from looking at
similar searches, I would suspect that we have several hundred
that apply to certified products. As I noted, there are over
100 for Windows. A few of the other products, the firewalls and
intrusion detection systems, the Oracle data base system has a
few as well. We know that there are several hundred
vulnerabilities for the bulk of certified products.
Mr. Clay. You also point out that software certified at the
highest level of the Common Criteria is subject to the same
worms and viruses as software that has not been certified. The
certification process is already a long and costly process.
What would have to be changed in the Common Criteria to address
these problems, and what would that do to the cost and time for
certification?
Mr. Spafford. Sir, the certification process is against the
documentation that is provided by the vendors and against a set
of specifications that have been put out by the groups that
have set the Common Criteria, and those do not include issues
such as resistance to malicious software.
There are architectural issues to the way that the code is
actually written that would need to be changed in the products
fundamentally. So, for instance, taking macros out of word
processing and spreadsheets, preventing e-mail programs from
automatically executing attachments are ways to stop viruses,
but they are not the only ways to stop those kinds of software
problems.
And those are not issues that are tested under the Common
Criteria. Those are architectural features that are actually
part of the product and the reason it is sold as it is.
Mr. Clay. Mr. Klaus, you indicate that you do not believe
that Common Criteria evaluation improves security. How would
you propose the government determine in the procurement process
that the software it buys is secure?
Mr. Klaus. I think the--in talking about the security
products, say, for example, the firewalls and the intrusion
detection systems, today there is no certification that I am
aware of that says, you know, take Gene Spafford's data base of
thousands of vulnerabilities and exploits and different ways
hackers get in today and evaluate whether a firewall or IDS
system stops all of these vulnerabilities and all of these
attacks.
So today there is no benchmark or operating system among
the security products to say which one is most robust against
these types of attacks; these are the known attacks much less
the unknown attacks that are continually evolving.
If we can just measure how good are security vendors
keeping up with the current pace of vulnerabilities, because
there is a lot. There are over 200 vulnerabilities, like I
said, every month, where we are tracking and need to keep
measuring the quality of the security vendors' products.
It is a little bit counterintuitive. I think if we look at
some of the commercial certification companies out there, they
have been able to hit the goals. When you look at the companies
that certify the antivirus companies, they meet 99.9 percent of
all viruses. They have been able to hit it. So they are testing
what is out in the wild, what are the latest things that are
happening, so they can quickly, at the end of the product,
measure did the security company keep in pace at the very end?
Did they hit the end result of what they said that they would
do for protecting against those threats?
And then the onus of having a very robust security product
and the processes are still left on the security vendor to
follow. And from a speed and cost perspective, because it is
only testing at the very end, did this thing catch all of the
hacker exploits, all of the worm exploits, all of the different
ways that systems get compromised, it is a much more
lightweight process. You can accomplish it in a month. It
doesn't take a year to go through that, and therefore the
turnaround is much faster.
Rather than trying to be completely overcomprehensive in
your evaluation of every detail and aspect of the security
design and architecture, I think that needs to be held to
probably the security vendor themself making sure that they end
up with that, because otherwise they themselves become part of
Gene Spafford's data base of vulnerable systems.
But, on the flip side, the most important thing in terms of
protecting the government: Can they stop these risks? I think
shrinking it down to a much more focused process would help
drive lower costs, faster speed and a much more accurate
measurement of is this product more secure or less secure for
the government.
Mr. Clay. Ms. Davidson, did you have something to add?
Ms. Davidson. I did. Actually I had a couple of responses
to that, one of them a personal anecdote. My company did look
at deploying one of the hottest new security products in a
particular sector, which is supposed to defend against certain
classes of application vulnerabilities. This is something, a
specialty firewall, that you would put out to protect yourself
against various types of attacks. It claimed to be the one of
the market-leading products. My hacking team broke it in less
than an hour using an attack that product was supposed to
prevent.
It is important that security products, because they are
the early warning system, have some type of independent
assessment of security worthiness.
I am also aware that people's intrusion detection systems
failed when Slammer was going around because of the
composability of the systems. They were running things back end
which themselves were not secure. I would certainly be open to
flexible ways of validating the security worthiness of security
products, but it is not all about feature function. It does one
no good to protect, allegedly protect, against certain classes
of attacks only to find that the security system itself is
badly flawed, and that in at least two cases has been our
experience.
Mr. Clay. Thank you.
Mr. Thompson, did you want to add something?
Mr. Thompson. I just wanted to point out that the security
evaluation process is designed to verify what a vendor claims,
but it does that--there is a very publicly available statement
of what the vendor claims. For example, if--you know, if the
toaster manufacturer says that--has this evaluated as a bread
storage device, it would be evaluated as a bread storage
device. And if the government wanted to buy toasters, and they
wrote a protection profile that specified toasters, the bread
storage device probably wouldn't meet the projection profile
for toasters. And somebody could write a security target for a
bread storage device, and it would be--you know, it would be
classified as a bread storage device.
The confusion--the CC process allows products to be
compared by specifying their security criteria in a semiformal
language that is easily comparable.
Mr. Clay. Well, the security issue sounds more like a
moving target, you know, as people come up every day with new
viruses, new worms, new ways to penetrate computers.
Mr. Thompson. Well, that is sort of a----
Mr. Clay. Can we win? Can we win the battle of securing
these computers?
Mr. Thompson. That is a different--finding patches and
fixing them is a very difficult process, very expensive
process, and I don't think that is the way we are going to win
in the end. There is certainly things we can do--to find a
patch is something we should do, as long as we have these
vulnerabilities, but the kind of software development that the
Common Criteria is encouraging is using sound engineering
principles and design life cycle processes.
And that is with the higher assurances like EL6 and 7,
encourage those kinds of things. In other words, you can't--if
you are going to evaluate EL7, you have to develop it in the
process of documenting. You have to formally prove that it
meets its security policies and things like that. And those
engineering principles have to be applied to the development
process. And we think that is a more--in the long run the only
way you are going to get secure products.
Mr. Clay. Mr. Klaus.
Mr. Klaus. I think there is--within the Common Criteria,
one of the issues that I think that the previous panel had
really pointed out repeatedly was that this is still an art
form in terms of finding the vulnerabilities, more R&D money
for automating the tools.
But at the end of the day, what we are finding is, is it
really requires subject matter experts to be able to--who
understand how to find buffer overflows, how to find heap
overflows, how to find--many of the techniques that hackers use
to break into the systems.
And what we find is, a lot of the approved testing labs
don't have that expertise, to find these kinds of
vulnerabilities; and from that perspective, we are not
measuring whether the systems have those types of
vulnerabilities. And I think if we can build a system that's
measuring for ``can the security products find and identify
attacks and stop them''--I think that's where Mary Anne
Davidson was pointing out that security products need to get
better at: on ``being evaluated,'' on ``can they identify the
attacks?''
Just as important as identifying the attacks, it is almost
more important for enterprises to make sure that we're also not
identifying false positives. This is where we falsely see, say,
legitimate traffic and identify it as, ``oh, here's an
attack,'' but the minute you do that, you start cutting off
real business transactions and so on, and then your security
product is no longer trusted; or you turn off that
functionality within the security product, and you are now less
secure.
And the other thing that needs to be tested is for invasion
techniques. A lot of the known hacker community has published--
there's lots of white papers on how to evade many of the
security products, and many of the security product vendors
behind that have not responded to those techniques. They are
still valid and still work.
And there's no, I guess, in the certification process,
anything that reflects how good they identify the attacks, the
false positives, the invasion techniques; and I think, to
answer another question, I think would be critical for security
companies to be measured on is the effect of ``zero-day''
exploits.
What I mean by zero-day is, when the worm comes out or a
virus comes out, the most impact, full-time, is within the
first 24 hours, in that it's spreading and nobody has
protection. All of the security vendors are trying to respond
to get the latest, ``What is that attack; OK, let's update our
security products.''
There's a concept of--within the security industry that
we're moving toward behavior-based security models, where I
don't have to take a fingerprint of every virus, every worm.
I'm actually looking at the behavior of that program so that if
something tries to compromise a system and acts to propagate
and format your hard drive and change your registries and other
things on the system, those are all bad behaviors, and it gets
flagged; and you could stop it without even knowing the--what
was the virus before you saw it.
And I think if we added some measurement to how good do
security products deal with zero-day threats, all you have to
do is test an old version--if it hasn't been updated, test
against a new threat. Did it stop it? If it did, great, you get
a point for that. If it didn't, you don't get a point and you
can start measuring across a lot of security parts out there.
Ms. Davidson. With all due respect, I think most of us
believe in defense and depth and that security cannot be
outsourced. If a vendor has a fault in their product, they
cannot outsource the remedy for that, even to intrusion
prevention.
For example, the customer comes to me and says they found a
fault in our software. I can't say to them, Do you have a fire
wall? Do you have an intrusion prevention system? Because if
you do, I won't fix it. They will have my head.
So I have to get it right the first time anyway. And if I
get it wrong, it will still cost me a million dollars to fix it
if it's on every single version of product on every operating
system.
Everyone needs to write better code. In order to write
better code, we need better tools. It's not just training,
because developers are human; they make mistakes. One mistake
and the hacker is in.
Mr. Spafford. I wanted to add, we have mentioned that we
need more research and tools. We need more personnel. This is
an area where we have a very small pool of expertise. But one
thing that would make a difference, I believe, is a matter of
accountability. And the sentiment expressed by Ms. Davidson
here has not been widespread enough within the industry, which
is, if there is a problem in the code, then the people who
wrote the code are held responsible.
Currently, when the government buys systems, if they have a
failure, then everybody rushes around and applies a patch and
then goes on as if nothing else had happened until the next
failure and the next patch.
I really believe that if there's some negative feedback to
the vendors involved, if they have a bad history of producing
software that isn't reliable, then perhaps that should be
figured into the next series of acquisitions. Perhaps there
should be a penalty applied to some vendors if they
consistently provide bad software. It's something worth
considering because simply encouraging them by buying the next
product cycle isn't resulting in the changes that we should be
seeing.
We are seeing vulnerabilities that have been known for 30
years to be security problems and bad practice; and we are
discovering that 50 percent of all the vulnerabilities that are
being reported today, 2 or 3 a week, are those bad practices
that are 30 years old and that my colleagues and I teach in the
very first few weeks for students to avoid. There should be
something back at--a negative pressure to have them start
paying attention to better practice.
Mr. Clay. Thank the panel for their responses.
Mr. Putnam. Could you give us an example of a 30-year-old
vulnerability?
Mr. Spafford. In the very introductory programing classes
we teach, we tell the students they should check the inputs.
For instance, if it's requested that a number be provided
between 1 and 10, we ask them to check that the value is
between 1 and 10. If they are asked to provide a character
string that is 10 characters long, then they should check to
make sure they aren't provided with one that is 11 characters
or 1,000 characters.
When we talk about buffer overflows or when you've heard
that mentioned by the panelists, that's a case where a program
was expecting 20 characters and was given 2,000 and there was
no check made to see that too many characters were provided.
That is something that has been known for 30 years to be a
problem. It has been exploited in many systems. We teach
against it, and it's still occurring and being discovered at
the rate of several a month.
Mr. Putnam. Ms. Davidson, Oracle began certifying products
very early in 1998. What led you to come to that conclusion and
how has it affected your business?
Ms. Davidson. We initially began doing evaluations,
actually the pre-Common Criteria days, we did the orange book
and IT section evaluations. Actually, we did four of them at
once on two of our products. We did that because one of our
core customer constituencies demanded it; at least we thought
they demanded it, as I testified previously.
They would occasionally wuss on the procurement
requirement--that's a technical term--but we kept doing them
anyway. We thought it was important. We found the benefits for
us were substantial for the reasons that I previously laid out.
I feel actually the cultural values--making security part
of a corporate culture has been the biggest value. I don't have
discussions or arguments about, are we going to hold the
release, because there's a security fault. Of course, we do
that. This is something that we are measuring on and it is
something we are held accountable on.
We certainly are not perfect. We have developers who have
committed the sin of buffer overflows or not checking input
conditions, but I would consider myself to be successful if I
could stomp buffer overflows in our time, but we need help to
do this. We spend a lot of money training people.
When I was in the Navy, there was an expression, ``To err
is human; to forgive is not Navy policy.'' People do make
mistakes in programming. If there are 21 conditions that have
to be validated, our developer checks 20 of them, the hacker
only needs to find that one.
If complexity is the enemy of security, so are manual
processes. The more that we can automate some of these checks
in addition to training people and holding them accountable,
the easier it will be for people to do the right thing. And
right now it is really hard, because you are only as good as
every single person checking every single possible programming
condition; and they're not perfect, and they never will be
perfect. It's been good for us as a business. And I don't think
the Common Criteria is a solution for all security ills, but I
think if people don't bake security into their development
processes, however we get there--just checking air conditioners
will also not make us secure; you need both.
Mr. Putnam. How would you respond to the toaster metaphor?
You know your company is committed to it, you follow through,
you are believers in Common Criteria. Mr. Spafford and to a
certain degree Mr. Klaus have laid out a series of arguments
why it will not get us where we hope that it will. How do you
respond to that?
Ms. Davidson. I think it is a great analogy on a lot of
levels.
The other counter-argument is, if you glue all the pieces
together, you may not get a secure house, but if you don't
start off with a secure foundation, you certainly will not get
a secure house.
Yes, evaluations do not make for perfect conditions. But
would you really want to plug your toaster in, even with the
cord, and have no idea how strong the building foundations
were, whether the engineers had done their jobs, whether they
had building inspectors in. You need to do a lot of things to
have secure software.
I think evaluations are part of the answer because it will
change the way people build software. It changed the way we
build software. I think have better validation. If we had
automated tools--most security faults are not faults in the
security mechanisms; they are as a result of bad programming.
If you have better automated checks for good programming
practice, you also will be able to add a level of robustness.
And the third piece I mentioned earlier is, many vendors,
despite our best efforts, don't deliver products that are
secure enough out of the box. We give our customers long lists
of things to do and to tweak to become secure. And most system
administrators never have enough hours in the day to do that.
You have to make it easy for people. You have to install
your product so that ideally people don't have to do anything--
like the Cuisinart--to have it operate securely. If you do
that, it not only lowers people-cost-of-operation and increases
their security, I think you will get resistance to some viruses
and worms that typically exploit lots of things that are left
wide open on your system, or things that are left lying around
in your system because a vendor shipped it and the customer
didn't know how to secure it.
Common Criteria is a strong necessity, but I would agree it
is not sufficient.
Mr. Klaus. I think from the house perspective, if you look
at how--I just went through the process of finishing a house.
The certification process and compliance is typically at the
end of the process. You have the government come out and look
at the house and make sure it's up to code and you meet that
criteria, or for a building or for--you are looking at the, at
the very end, did the House meet all the necessary standards.
And some of the important--and it's looking for the critical
issues. Are sprinklers in place, etc.
What you don't see in the certification process--and this
is where I think we are failing--is, the opposite is happening
in the Common Criteria where if you had a document--as the
architect, the designer of the house had to sit there and as it
goes through the process add another year--I mean, it took a
long enough time to finish the house. If you add another year
to building the house, to make sure that everything was
documented, and here was all--I trust my architect to make sure
it's designed to be built strongly.
The government shouldn't have to go in there and say, did
you use all the metrics to make sure it's going to stand, and
then at the end the government checks to makes sure the most
important issues are addressed and certified. And to the
extent--if we could move the Common Criteria more to, did the
important issues get addressed?
And I think you could look at it, hey, a lot of these
applications, especially business applications, are very
complex. Many ways--many lines of code, etc. But if you
actually identify what are the most common ways that hackers,
worms, viruses hack into a system, the majority of the risk is
at the network protocol code. You know, if you look at, why did
Blaster get into the operating system, well, it was because
there was an RPC service running on every Windows box that had
this vulnerability.
You can say there's millions and millions and millions of
lines of code within operating systems and these business
applications, but the most important thing is to look at, what
are the things that are exposed at the network level? That
tremendously reduces what you have to evaluate.
We do a lot of security penetration tests, a lot of
security assessments trying to figure out how would a hacker
break into a system, and we always start at the network layer.
And I think if the certification process looked more at--the
same way that the hackers, the worms and viruses looked at, how
does somebody break into the system, you'd start saying, OK, do
you want to check the doors and the windows? You don't want
to--I mean, you don't try to evaluate every wall and floor, the
whole house. You evaluate the areas that hackers get into.
And that's kind of intuitive, but if we focused on the
bigger issues measuring whether you have a good security
product or not; less on, did the overall process get followed,
because right now it's not helping us find the buffer overflows
and other things within the product at the end of this
certification process.
Mr. Putnam. Dr. Spafford, you started this metaphor, and I
would ask for you to talk a little bit about what the better
alternative is. Software assurance, how do we get there if it's
not Common Criteria?
Mr. Spafford. I didn't mean my comments to mean that Common
Criteria is not a value, because I believe it is. It provides
guidance as to how go about building a quality product. But
it's building that quality product that is the key to what we
are talking about.
It's not simply a matter of security. We want to have
greater trust in our systems, but we also want it to be
reliable in the face of failure, unexpected circumstances.
It appears, for instance, that the blackout that occurred
in the East Coast was as a result of unfortunate circumstances
happening at once, without sufficient capacity and reserve to
make up for the failure. We don't want that to occur with our
computer systems either.
That means going back and looking at fundamental
assumptions that are made on how we build the systems. What are
the features that we really want? How is it being built? Is it
being built using good tools and by people who understand the
technology? Are they putting in more features than are really
necessary, which I believe is the root cause of a number of the
problems that we see. Is the documentation in the interface?
Are those two items put together in such a way that the average
user is able to understand how to use the system and how to
configure it?
Again, I do not believe that is the case. The average user
currently is very often someone at home who doesn't understand
what a firewall is or a virus is or what it means to have their
system connected all the time.
Then we have to have better testing tools and some known
reasons to test, some known test sets to work against. We have
to be able to test in real environments, so that if we are
going to deploy something in a large-scale system, we have to
have testbeds to do that; and again, we have to have the people
trained to do that.
And last of all, we have to have a mechanism so that we
understand if we need to apply the technology to new arenas,
how we go about going back in the process and changing the
technology rather than simply reusing the old technology
because that's what we have a large investment in. We should be
using the most appropriate tools for the tasks at hand.
What has happened over the last 30 years for computing, if
we look, there's been incredible strides from mainframes and
small networks to where we are now with global, international
activity with our systems. We don't even know where some of our
software comes from because of the international trade and
development that goes on.
We spent those 30 years trying to make the technology work,
and I think we have done a really admirable job of that. So
much of our society, so much of our dominance in the world has
come about through our ability to create good technology. But
now we have to change our mind-set to think about how to most
appropriately use that technology and how to make it safe, and
that means really taking a step forward, leaving behind some of
the technologies of the past.
So again, to summarize, it's not a simple step. It's going
to be a whole number of steps throughout the life cycle of
building and designing software. And to revisit Mr. Klaus'
comments about the architect, well, the architect has been
through many years of professional training. They probably
served an apprenticeship with a master architect to understand
what they're doing. And if they have designed his house, and it
ends up getting built and there's no doors and it collapses
afterwards, he has some recourse. And it's possible that
architect will not be able to sell a design again in the
future.
We haven't done that in the software arena. We need to
start thinking in terms of how we're going to protect our
future. Are we going to continue to reward bad performance?
So it's a long answer--I apologize--but it's a very
multifaceted problem.
Mr. Putnam. The $64,000 question: Should we expand Common
Criteria to civilian agencies?
Mr. Spafford. I believe that, on balance, that should not
be mandatory. As a voluntary step, it may be good, but
mandatory, it will not solve the basic problems. There are
certified products that won't work as required.
The process is not easy to understand. The Common Criteria
standard document is 700 pages long, and so many of the people
are going to be buying and deploying these systems who won't
understand what the certification means, which is why I used
the analogy of the toaster. The average consumer won't
understand what that means.
It can help to get some vendors to pay more attention, but
I believe that the additional overhead, time and costs that
were discussed by the other panelists are probably
counterproductive to government's needs. I believe there are
other steps that should be taken first.
Mr. Klaus. My answer would be, until the, at least the 3
years we talked about, better measurement of what you are
trying to do with the Common Criteria is met, meaning, is this
product actually providing better protection against the known
threats and making process later, because if it takes a year to
get our products out the door to help the government, they're
going to be a year behind the commercial sector. And the cost
of it is--I think overall, the cost is expensive, so startups
have--will have a hard time entering into the government
sector.
But most importantly, if the cost was moving toward making
the products better, I'd be in favor of it. Today there's very
little value in what it is today.
Mr. Putnam. Ms. Davidson.
Ms. Davidson. If it's not too expensive and doesn't take
too long to do.
With all due respect to Mr. Klaus' company and their fine
products, we have more complex products. We get certificates
out within 6 months of the production release, and we release
major versions of product every year to 18 months. It is cheap
compared with the alternative.
We are already paying for bad security. I believe that it
should be extended at least--clearly, to entities who have a
national security focus. And if the Department of Homeland
Security is not doing national security, what is it that
they're doing?
As I mentioned earlier, there are other things we can do,
but unless we fundamentally, as an industry, change the way
that we build product, nothing will ever change. And this is
the government's last chance on this. If we abandon information
assurance efforts and go only to a testing approach, you will
never know whether someone developed good product.
Testing alone, while I think an important add-on on top of
Common Criteria evaluations, also will not solve the problem.
Mr. Thompson. I think I agree with Gene that we have to
encourage good software development, good software design, and
encourage companies to develop products that are safe in the
beginning; and not just throw them on the market and let the
rest of the world find the bugs one at a time or let the
hackers find the bugs. They need to produce good software and
need to be held accountable when they don't.
And the Common Criteria approach to evaluation encourages
vendors to do that. It is not designed to find bugs in a
particular release of a product, but designed to encourage
vendors to use good software and build that house according to
good architectural principles in the beginning and not to find,
you know, where the beams have been left out or where the small
beams were used.
Expanding the market for evaluated products would encourage
that, send a signal to industry that the government is serious
about good software engineering and development; and the
products should be, you know, secure from the get-go. And
anything you can do to allow--make vendors accountable to
their--for putting out bad software would further the
government's ability to buy good software. Everyone would have
better software available if vendors were held accountable.
Mr. Putnam. Thank you, Mr. Thompson.
I want to thank all of our witnesses today, particularly
the second panel, for their efforts in helping us to better
understand this very complicated issue.
Gaining assurance that the software the government buys to
protect itself actually can do the job is an important goal.
I also want to thank Mr. Clay and Ms. Watson for their
participation. In the event that there may be additional
questions that we did not have time for, the record shall
remain open for 2 weeks for submitted questions and answers.
With that, the subcommittee stands adjourned.
[Whereupon, at 12:15 p.m., the subcommittee was adjourned.]