Electronic Benefits Transfer: Use of Biometrics to Deter Fraud in the
Nationwide EBT Program (Letter Report, 09/29/95, GAO/OSI-95-20).
Pursuant to a congressional request, GAO examined options for reducing
fraud in the electronic benefits transfer (EBT) environment, focusing on
the use of biometrics to verify recipients' identities and reduce the
potential for fraud.
GAO found that: (1) some states and localities have used biometrics to
deter fraud in their welfare benefit programs and have realized major
cost savings by requiring program applicants to submit to electronic
fingerprinting during enrollment; (2) electronic fingerprinting may be
the most viable option for deterring fraud in an EBT environment, since
it has been universally accepted and has proven to be reliable; (3)
fingerprinting benefit applicants during enrollment would eliminate
losses related to duplicate benefits payments; (4) fingerprint
verification during benefit disbursement would directly link withdrawals
to recipients and help resolve potential losses and increased costs; and
(5) the use of biometrics in an EBT environment should be tested before
nationwide expansion of the EBT program.
--------------------------- Indexing Terms -----------------------------
REPORTNUM: OSI-95-20
TITLE: Electronic Benefits Transfer: Use of Biometrics to Deter
Fraud in the Nationwide EBT Program
DATE: 09/29/95
SUBJECT: Fraud
Electronic funds transfer
Welfare benefits
State-administered programs
Program abuses
Food stamp programs
Losses
Welfare recipients
Public assistance programs
Cost control
IDENTIFIER: USDA Electronic Benefit Transfer System
AFDC
Aid to Families with Dependent Children Program
WIC
Special Supplemental Food Program for Women, Infants, and
Children
FBI Integrated Automated Fingerprint Identification System
Los Angeles County (CA)
**************************************************************************
* This file contains an ASCII representation of the text of a GAO *
* report. Delineations within the text indicating chapter titles, *
* headings, and bullets are preserved. Major divisions and subdivisions *
* of the text, such as Chapters, Sections, and Appendixes, are *
* identified by double and single lines. The numbers on the right end *
* of these lines indicate the position of each of the subsections in the *
* document outline. These numbers do NOT correspond with the page *
* numbers of the printed product. *
* *
* No attempt has been made to display graphic images, although figure *
* captions are reproduced. Tables are included, but may not resemble *
* those in the printed version. *
* *
* A printed copy of this report may be obtained from the GAO Document *
* Distribution Facility by calling (202) 512-6000, by faxing your *
* request to (301) 258-4066, or by writing to P.O. Box 6015, *
* Gaithersburg, MD 20884-6015. We are unable to accept electronic orders *
* for printed documents at this time. *
**************************************************************************
Cover
================================================================ COVER
Report to the Honorable
Kenneth E. Bentsen, Jr., House of Representatives
September 1995
ELECTRONIC BENEFITS TRANSFER - USE
OF BIOMETRICS TO DETER FRAUD IN
THE NATIONWIDE EBT PROGRAM
GAO/OSI-95-20
Use of Biometrics in Proposed EBT Program
(600376)
Abbreviations
=============================================================== ABBREV
AFDC - Aid to Families With Dependent Children
AFIS - Automated Fingerprint Identification System
ATM - automated teller machine
EBT - electronic benefits transfer
FBI - Federal Bureau of Investigation
GAO - General Accounting Office
IAFIS - Integrated Automated Fingerprint Identification System
IEI - Invitation for Expressions of Interest
INS - Immigration and Naturalization Service
OIG - Office of Inspector General
OSI - Office of Special Investigations
PIN - personal identification number
RCED - Resources, Community, and Economic Development Division
USDA - U.S. Department of Agriculture
WIC - Women, Infants and Children
Letter
=============================================================== LETTER
B-261923
September 29, 1995
The Honorable Kenneth E. Bentsen, Jr.
House of Representatives
Dear Mr. Bentsen:
In 1993, the National Performance Review recommended that the federal
government consider the potential for providing all payments to
individuals by using electronic, rather than paper, payments.\1 In
1994, the Federal Electronic Benefits Transfer Task Force--consisting
of officials from the Office of Management and Budget, the Department
of Health and Human Services, the U.S. Department of Agriculture
(USDA), and the U.S. Department of the Treasury--reported that each
year federal and state programs deliver almost $500 billion in cash
benefits and food assistance and that at least 12 federal and state
benefit programs could use electronic benefits transfer (EBT) to
replace the current paper benefit delivery methods.\2 The task force
estimated that over $110 billion in annual cash benefits and food
assistance could be delivered with EBT, including such benefits as
food stamps, social security, and federal pensions.
The task force determined that an electronic system would reduce the
cost of benefit delivery, strengthen the management of program funds,
and reduce fraud. Under such a system, federal, state, or local
government agencies would issue access cards (similar to credit
cards) and personal identification numbers (PIN) to recipients who
could obtain benefits through automated teller machines (ATM) and
point-of-sale terminals.\3 EBT is already assisting USDA by providing
data that can be analyzed by computer programs to target stores
trafficking food stamp benefits and identify individuals who frequent
stores suspected of trafficking.
Because of the significant federal funding involved in government
programs providing benefits, we examined various options for
providing additional security to deter the potential for fraud in an
EBT environment. We focused on the use of biometrics--automated
methods to measure a physical characteristic or personal trait--to
verify a recipient's identity and reduce the potential for fraud.
Because of that potential for fraud, on August 11, 1995, you
requested that we report to you on our work. Appendix I provides an
overview of the biometric technologies we reviewed--fingerprints,
hand geometry, retina scan, voice verification, and signature
verification.
--------------------
\1 From Red Tape to Results, National Performance Review (Washington,
D.C.: Sept. 1993).
\2 Creating a Benefit Delivery System That Works Better & Costs Less,
Federal EBT Task Force (Washington, D.C.: May 1994).
\3 A point-of-sale terminal is a device placed in a merchant location
and connected to a bank's system by telephone lines. It is designed
to authorize, record, and forward electronically the payment for each
sale as it occurs.
RESULTS IN BRIEF
------------------------------------------------------------ Letter :1
Some states and municipalities have used biometrics to deter fraud in
their social welfare benefit programs. They have realized
substantial cost savings by requiring program applicants to submit to
electronic fingerprinting as part of the enrollment process. Using
this method, they have denied benefits to individuals who attempted
to receive duplicate benefits. For example, the Los Angeles County
Department of Public Social Services reported savings of $14 million
solely attributable to such a system from June 1991 through July
1994. The U.S. Secret Service supports the use of fingerprint
identification in the benefit enrollment process and commented
favorably on the success of the Los Angeles County project.
Electronic fingerprint identification offers a promising solution for
deterring fraud in both the enrollment and disbursement phases of the
government's proposed EBT program. Of the biometric identification
systems that we reviewed, we selected fingerprinting as the one most
viable for verifying a recipient's identity in an EBT environment.
We selected it because of (1) its universal acceptance as a positive
means for identity verification and (2) its extensive history of
reliability in the law enforcement arena. Fingerprinting benefit
applicants during the enrollment phase would eliminate losses related
to applicants' applying for duplicate benefit payments under
different names and deter others so inclined. Such verification in
the disbursement phase would directly link withdrawals to the
recipients and effectively resolve the issue of potential losses and
increased costs. Federal regulations would limit consumer liability
when lost or stolen EBT cards are misused. However, the issue
concerning who will be liable for losses over the consumer limit,
which could be extensive, is still unresolved.
The effectiveness of an EBT program secured by biometric
identification to deter fraud should be tested in an EBT environment
before the program is expanded nationwide. The development and
testing of a biometric system may delay the task force's proposed
1999 implementation date for the EBT program and would increase the
program's initial cost. However, the long-term benefits of a
biometric system would contribute to a more fiscally sound and secure
EBT program.
This report contains a recommendation aimed at reducing potential
fraud and abuse in the nationwide EBT program.
BACKGROUND
------------------------------------------------------------ Letter :2
The Federal Electronic Benefits Transfer Task Force has proposed that
the federal government use EBT to disburse such benefits as social
security; railroad retirement; federal civilian retirement; military
pensions; food stamps; Aid to Families With Dependent Children
(AFDC); and Women, Infants and Children (WIC) in all states by 1999.
In fiscal year 1994, these programs disbursed about $433 billion in
federal benefits through such delivery methods as electronic funds
transfer (direct deposit) and Treasury checks. For example, in
fiscal year 1994, the federal share of the food stamp program was
about $24 billion, disbursed to over 27 million recipients in mainly
food stamp coupons. Various law enforcement officials estimate the
losses from fraud in existing programs, such as food stamps, to be up
to 10 percent annually. We have not verified these loss estimates.
Because of the multiplicity of government agencies that could
participate in the EBT program, the task force will develop and
oversee the national EBT policy, manage EBT prototype projects, and
coordinate budget requests related to implementation and operation of
a nationwide EBT program. The U.S. Department of the Treasury will
manage the federal government's financial operations associated with
EBT. In March 1995, Treasury issued an Invitation for Expressions of
Interest (IEI) to acquire EBT services for the Southern Alliance of
States\4 as a prototype. Treasury expects to award a contract in
October 1995. According to a member of the Federal EBT Task Force,
the Treasury IEI strongly encourages financial institutions to
recommend the use of innovative technologies, such as biometrics,
when formulating a response to the IEI. In addition, he told us that
the task force supports the most secure and cost-effective measures
or technologies to safeguard EBT systems and is looking to the
banking industry or existing commercial banking infrastructure to
take the lead with innovative technologies.
Numerous methods exist to afford different levels of security to
safeguard an EBT system. The methods range from magnetic stripes
encoded with various information to such sophisticated biometric
techniques as fingerprints, hand geometry, retina scan, voice
verification, and signature verification. However, Secret Service
investigations have shown that the less sophisticated levels of card
security--such as those cited in Treasury's IEI, including magnetic
stripes and holograms--have been counterfeited. Individuals have
counterfeited credit/debit cards themselves and easily transferred
the information encoded on magnetic stripes from one card to another.
Of the physical characteristics and personal traits that are used for
biometric verification, law enforcement agencies have used
fingerprints most extensively. For almost 100 years, law enforcement
has used fingerprints to identify criminals both upon arrest and
after comparing crime scene fingerprints with already established
criminal fingerprint files. Advances in electronic fingerprint
identification have resulted in positive identifications in criminal
cases once left unsolved after using manual fingerprint search
methods. Some police jurisdictions are now using live scan
fingerprint capture\5 to fingerprint individuals they arrest.
During benefit enrollment, live scan fingerprint capture allows for
identity verification through database search. During disbursement
at ATM or point-of-sale terminals, this technology allows for
self-verification by using a fingerprint reader to compare a live
scanned fingerprint with the same print encoded on an EBT card.
Figure 1 depicts the use of electronic fingerprint verification
during benefit enrollment (database search) and benefit disbursement
at point-of-sale terminals (self-verification).
Figure 1: Fingerprint
Verification During Benefit
Enrollment and Benefit
Disbursement at Point- of-Sale
Terminals
(See figure in printed
edition.)
With a fingerprint-secured EBT card, a program administrator could
link the responsibility for use of the card to the recipient and, if
fraud was alleged, have the information needed to determine a future
course of action. A fingerprint-secured card could not be used by
anyone other than the authorized recipient of the entitled benefits.
--------------------
\4 The alliance is currently a coalition of nine southern
states--Alabama, Arkansas, Florida, Georgia, Kentucky, Mississippi,
Missouri, North Carolina, and Tennessee--that joined with the task
force to develop the specifications for a prototype national EBT
system.
\5 This technology involves using equipment to directly scan a finger
and store or transmit the data electronically.
AVAILABLE TECHNOLOGY OFFERS
POTENTIAL TO REDUCE FRAUD
------------------------------------------------------------ Letter :3
In December 1994, we reported that the state EBT systems and pilot
projects we reviewed, including those used to distribute food stamps,
have not eliminated fraud.\6 For example, the first major fraud
investigation by USDA's Office of Inspector General (OIG) involving
EBT in Pennsylvania found that a small sandwich shop had conducted
over $151,000 in fraudulent EBT transactions over a 2-year period.
These transactions accounted for 76 percent of the shop's total EBT
dollar volume, and 173 food stamp recipients were convicted for
selling their EBT benefits. During another investigation in
Pennsylvania, OIG agents found that a retail establishment had
illegally obtained 79 EBT cards along with the recipients' PINs. In
addition, in February 1995, the USDA Inspector General testified that
his office, through analysis of EBT data, had identified about 7,500
food stamp recipients who appear to have sold almost $2 million of
their benefits in Baltimore, Maryland, between August 1992 and
February 1994.\7
Available technology, including biometrics, exists to help reduce the
potential for fraud in EBT programs. For example, some state and
local governments have started to use the Automated Fingerprint
Identification System (AFIS) to deter fraud in social service
programs during the enrollment stage by identifying subsequent
requests for duplicate benefits by an individual. The most widely
publicized effort--the Los Angeles County Department of Public Social
Services--began using live scan capture of index fingerprints to
enroll applicants for participation in one of its social welfare
programs in 1991.
Los Angeles County has the second largest county-operated welfare
department in the United States, with a $3-billion budget. The
county, using AFIS technology, piloted a fingerprint
project--Automated Fingerprint Image Reporting and Match--with its
General Relief program, which provides financial assistance to
indigent persons and emergency assistance to individuals and families
in temporary need. The county spent $9.6 million to purchase the
hardware and software for the system and reported $14 million in
savings solely attributable to the project from June 1991 through
July 1994.
Of this amount, according to the county, it realized $5.4 million in
savings during the first 6 months the project was used as a result of
terminating over 3,000 approved cases and denying over 240 cases for
failure to comply with its fingerprinting requirements. In an
October 1994 preliminary evaluation of the project, the audit firm of
Ernst & Young reported that (1) only 3 percent of the client
population experienced negative feelings about being fingerprinted,
(2) the system had not increased the amount of time clients wait in
line to apply for benefits, and (3) the system can be replicated
effectively and extended to other programs and locales. In addition,
the California counties of Alameda and Contra Costa and the city of
San Francisco have implemented fingerprint verification for use in
some of their social benefit programs. These jurisdictions share
data, and officials believe that the system has discouraged
applicants from applying for duplicate benefits in more than one
county.
The U.S. Secret Service has commented favorably on Los Angeles
County's efforts because of the benefits that a fingerprint
enrollment verification system offers. These same officials maintain
that failure to use the available fingerprinting technology to deter
fraud in the initial enrollment phase of the program may open the
entire system to fraud and abuse.
--------------------
\6 Food Assistance: Potential Impacts of Alternative Systems for
Delivering Food Stamp Program Benefits (GAO/RCED-95-13, Dec. 16,
1994).
\7 Testimony by the Inspector General, USDA, before the House of
Representatives, Committee on Agriculture (Feb. 1, 1995).
UNRESOLVED ISSUES RELATING TO
IMPLEMENTING A FISCALLY SOUND
NATIONWIDE EBT SYSTEM WITH
ENHANCED SECURITY
------------------------------------------------------------ Letter :4
Action is necessary to help ensure the fiscal stability of a
nationwide EBT system and enhance its security. Such action includes
(1) resolving the issue of potential losses and increased costs
related to consumer liability protection in the event of the misuse
of lost or stolen EBT benefit cards and (2) pilot testing an EBT
system secured by biometric identification before expanding the
program nationwide.
ADDITIONAL COSTS COULD BE
ASSOCIATED WITH CONSUMER
LIABILITY PROTECTION DUE TO
FRAUD
---------------------------------------------------------- Letter :4.1
Regulation E, under 12 C.F.R. part 205, implements the Electronic
Fund Transfer Act, 15 U.S.C. sections 1693-1693r (1994). The act
and regulation (1) cover any electronic fund transfer initiated
through an ATM and point-of-sale terminal, automated clearinghouse,
telephone bill-payment system, or home banking program and (2)
provide rules that govern these and other electronic transfers.
The act limits a consumer's liability for unauthorized use of credit
and ATM cards, telephone bill-payment systems, or home banking
programs. A consumer's liability is limited to $50 if the consumer
notifies the account-holding institution within 2 days after learning
of a loss, theft, or unauthorized use. The card issuer may then
assume the losses in excess of $50. Effective February 28, 1994, the
Federal Reserve Board amended Regulation E to apply the same consumer
liability to EBT programs established by federal, state, or local
government agencies. These entities must comply with the regulations
by March 1, 1997.\8
Some of those who commented on the proposed amendment to Regulation E
noted that the liability protection may result in additional costs in
an EBT program. These individuals pointed out that, presently,
financial institutions can control their costs from misused EBT
services by selecting the customers to whom they offer the services.
However, government agencies must accept all who qualify for the
benefit program. In addition, if a customer of a financial
institution is suspected of engaging in fraud, the institution can
terminate the account relationship. With the consumer liability
protection of Regulation E, EBT recipients engaged in fraud could
sell their cards and PINs, report the card and PIN as lost or stolen,
obtain a new EBT card and replacement benefits, and be liable for
only $50 in lost benefits. Many times that amount could be obtained
fraudulently before the card was canceled. Such schemes would be
difficult to police without evidence of intentional fraud on the part
of the recipients. Directly linking the responsibility for
withdrawals to individual recipients by using fingerprint
verification could effectively eliminate this type of fraud.
The possibility of substantial fraud losses related to Regulation E
implementation has been estimated by some of those responding to the
proposed amendment to be between $164 million and $986 million
annually. A few of the estimates were based on agency experience
with the replacement of lost or stolen cards in EBT programs, but
most of the cost estimates were based on loss and fraud experience
under such existing paper-based benefit programs as food stamps.
Some of those commenting on Regulation E also noted that private
sector institutions handle losses related to the Regulation E
customer liability limitations by spreading the losses over their
entire customer base in the form of increased fees or reduced
interest paid. Some explained that government agencies cannot do the
same; therefore, losses would have to be paid out of tax revenues or
by reducing benefits.
However, neither the Federal EBT Task Force nor the Treasury has
resolved who or what entity would be responsible for absorbing the
costs above the $50 consumer liability limit that could result from
the misuse of lost or stolen cards. According to the Federal EBT
Task Force report, "[A]ll agencies are concerned about assuming
liabilities of undetermined value."
--------------------
\8 Pending legislation would exempt state or local government
programs from Regulation E. See H.R. 4, 104th Cong., 1st Sess. µ
802 (1995) and S. 131, 104th Cong. 1st Sess., µ 1 (1995).
PILOT TESTING NEEDED TO
ASSESS BENEFIT OF BIOMETRICS
IN EBT ENVIRONMENT
---------------------------------------------------------- Letter :4.2
Pilot testing of biometrics in an EBT environment before the EBT
program is expanded nationwide would allow for an assessment of its
practicality and effectiveness in combatting fraud in new EBT
systems. For example, pilot testing of self-verifying live scan
fingerprint readers in conjunction with stand-alone processing
capabilities at ATMs and point-of-sale terminals would also help
evaluate fraud deterrence at the disbursement level. Pilot testing
would provide a means for determining the reliability and accuracy of
the equipment in an actual ATM/point-of sale environment and serve to
identify other possible problems.
User-friendliness and customer satisfaction could also be assessed
during pilot testing of a fingerprint verification system. Such
testing could also assist in determining the best placement of the
equipment to make it easily accessible for participants.
CONCLUSIONS
------------------------------------------------------------ Letter :5
EBT alone does not effectively deter fraud in the delivery of food
stamp benefits. Thus, an EBT program without the enhanced security
of biometric verification raises a genuine concern about the
potential for increased program costs and losses. The concern
increases with the proposal to expand EBT into other federal, state,
or local government programs involving billions of dollars--such as
AFDC, WIC, social security, and federal retirement benefits--and with
full implementation of Regulation E.
Due to the universal acceptance of fingerprints as a means for
verifying identities, its extensive history of reliability in the law
enforcement arena, and successes with AFIS technology, we believe
fingerprint verification is the biometric form that offers the
greatest potential for success and acceptance in securing EBT systems
from fraud. Further, an EBT system with fingerprint verification
would effectively negate the cost/loss concerns raised by Regulation
E implementation.
Development and testing of an EBT system with biometric safeguards
would increase the cost, largely from purchasing hardware and
software, and time to implement the nationwide system. Yet the
development and testing are necessary to ensure that the future
system is practical and not beset with the problems of fraudulent
usage. Further, such development, testing, and ultimate use would
reduce losses to the EBT program from fraud and abuse.
RECOMMENDATION
------------------------------------------------------------ Letter :6
We recommend that the Secretary of the Treasury develop, if feasible,
a biometric verification system, such as electronic fingerprinting,
for use in an EBT environment. The use of biometrics could be
assessed in a limited area, such as the Southern Alliance of States,
and prior to expansion of the EBT program nationwide.
AGENCY COMMENTS
------------------------------------------------------------ Letter :7
The Department of the Treasury, in commenting on a draft of this
report, agreed that biometrics may provide a cost/benefit advantage
in the eligibility phase of the EBT program. Although Treasury did
not dismiss the use of biometrics in the delivery of EBT benefits, it
expressed reservations about the immediate use of biometrics in
delivery, noting that such use would require considerable testing and
study. Treasury also cited the existence of limited data concerning
both the use of biometrics and the possible impact of Regulation E,
the EBT program's need to be accessible and cost effective, and a
concern to preserve the dignity of recipients. Treasury further
stated that to enhance card security, it is requiring that EBT cards
include countermeasures against counterfeiting. (See app. II.)
We agree that testing and study are needed to resolve technological
and policy issues. However, because fraud other than counterfeiting
persists in EBT pilots around the country, we believe that, if
feasible, the evaluation should be completed before full
implementation of the nationwide EBT program as envisioned by the
Federal EBT Task Force in 1999. Further, we are optimistic that
commercial and banking entities that use debit and credit cards will
implement biometric safeguards, such as fingerprinting, to protect
their customers and themselves. This will help create a convenient,
cost-effective EBT environment in which most recipients should not
feel singled out.
METHODOLOGY
------------------------------------------------------------ Letter :8
To obtain the information in this report, we examined various
biometric systems available for identification
verification--fingerprints, hand geometry, retina scan, voice
verification, and signature verification. We interviewed officials
of federal agencies that use such technologies, including the Federal
Bureau of Investigation (FBI), the U.S. Secret Service, the U.S.
Immigration and Naturalization Service, and the Federal Bureau of
Prisons. We also interviewed state and law enforcement officials in
California and various biometrics vendors. We visited sites where
federal, state, and local governments and private entities operated
various types of biometric identification programs. We attended
conferences on biometrics at the FBI, Quantico, Virginia, and in
Washington, D.C.
In addition, we reviewed numerous documents including the Federal
Reserve Systems' final rule on Regulation E, the Federal Electronic
Benefits Transfer Task Force's May 1994 report, a March 1994 U.S.
Department of the Treasury IEI for EBT services, and an October 1994
preliminary evaluation by Ernst & Young of the Automated Fingerprint
Image Reporting and Match system being used by the Los Angeles County
Department of Public Social Services. We also reviewed various
reports issued by the Department of Justice's Bureau of Justice
Statistics, the FBI, the National Security Agency, and Sandia
National Laboratories, as well as materials provided by biometrics
industry vendors. We conducted our review from November 1994 through
June 1995.
---------------------------------------------------------- Letter :8.1
We will send copies of this report to appropriate congressional
committees, the Secretaries of Agriculture and Treasury, and the
Director of the Office of Management and Budget. If you have
questions concerning these issues, please contact me or Assistant
Director Houston Fuller of my staff at (202) 512-6722.
Sincerely yours,
Richard C. Stiener
Director
SELECTED BIOMETRIC TECHNOLOGIES
AND THEIR USES
=========================================================== Appendix I
Biometric technologies use an automated method to measure a physical
characteristic or personal trait to verify an individual's identity.
Some biometric technologies currently being marketed and used include
(1) fingerprints, (2) hand geometry, (3) retina scan, (4) voice
verification, and (5) signature verification. Federal, state, and
local governments and the private sector have used these technologies
to ensure the security of computers, facilities, welfare benefits,
and credit/debit cards. An overview of these technologies follows.
FINGERPRINTS
--------------------------------------------------------- Appendix I:1
Fingerprints have been used by law enforcement agencies for almost
100 years to identify criminals both upon arrest and after comparison
of crime scene fingerprints with already established criminal
fingerprint files. Fingerprints provide both a permanent and
positive identification system for law enforcement and civilian
purposes. Although two fingerprint patterns may be similar, no two
fingerprints have ever been found to contain identical individual
ridge characteristics. These characteristics are present on normal
hands and feet some months before birth and are constant, except for
accidental damage, until decomposition after death.
THE HENRY FINGERPRINT
CLASSIFICATION SYSTEM
------------------------------------------------------- Appendix I:1.1
The Henry System, for years the predominant fingerprint
classification system, assigns each finger to one of three primary
fingerprint pattern types: arches, loops, or whorls.\9 In this
system, the fingerprints are represented as a unit rather than as
individual fingers by assigning to each 10-print set an alphanumeric
designation reflecting the pattern characteristics of all 10 fingers.
This classification system was a major step forward in the use of
fingerprints because it enabled fingerprint forms bearing differing
patterns to be placed in a certain order, thus enabling the search
area to be minimized. Figure I.1 shows the primary fingerprint
patterns.
Figure I.1: Primary
Fingerprint Patterns
(See figure in printed
edition.)
Source: FBI
--------------------
\9 The Henry System, credited to an Englishman, Sir Edward Henry,
became operational at Scotland Yard in 1901.
BENEFITS OF THE AUTOMATED
FINGERPRINT IDENTIFICATION
SYSTEM (AFIS)
------------------------------------------------------- Appendix I:1.2
With the increasing size of fingerprint databases, manual searches
under the Henry System have become too time consuming; and the
identification of latent prints,\10 even more difficult. Thus, in
the early 1980s, U.S. law enforcement agencies began using the
Automated Fingerprint Identification System (AFIS), researched at the
National Bureau of Standards\11 under the sponsorship of the FBI
during the early 1960s. According to law enforcement officials, the
system has been accurate between 98 and 100 percent of the time in
searching and matching fingerprints. AFIS uses computers to scan and
digitize fingerprints by automatically creating a spatial geometry or
map of the unique ridge patterns of the prints and translating the
spatial relationship into a binary code for the computer's searching
algorithm. Figure I.2 depicts the plotting of fingerprint minutiae
by AFIS.
Figure I.2: AFIS Plotting of
Fingerprint Minutiae
(See figure in printed
edition.)
Source: FBI
The following briefly describes the success experienced by various
law enforcement agencies with AFIS.
One large urban police department that has used AFIS since 1983
found that the system can conduct a "cold search"\12 in a
database of 340,000 10-print cards in about 1 minute. The same
police department can run a 10-print card against its latent
file of 6,000 prints in about 40 seconds and realize a positive
identification about 22 percent of the time, compared with 8
percent prior to the system's implementation. In the first year
of operation, the department conducted 5,514 latent print
searches on its AFIS system and made 1,001 identifications, an
identification rate of over 18 percent. The department cleared
816 of those cases compared with 58 cases cleared the previous
year on the basis of latent print identifications.
According to California law enforcement officials, they have
realized a 98-percent accuracy rate for matching a single
10-print card with the 8.5-million 10-print cards in the state's
AFIS system and a 15-percent rate on latent print searches.
Several noteworthy cases were solved through searches conducted
on latent prints in California's AFIS system. In 1985, three
murder suspects were arrested after a single thumbprint was
found on a vehicle owned by one of the victims. The vehicle was
found abandoned and burning, with little chance of obtaining
other evidence. In 1987, approximately a decade after a man was
murdered and his son paralyzed by an assailant's bullets, the
state identified a suspect with a single thumbprint found on a
kitchen window at the crime scene. And in 1988, nearly 21 years
after a murder was committed, the state identified a suspect by
searching latent prints found at the time of the murder.
California is a member of the Western Identification Network, a
multistate organization formed in 1989 to exchange automated
fingerprint information. As a result, California exchanges
information with the member states of Nevada, Oregon, Idaho, Utah,
Wyoming, and Montana, and three associate member states, as well as
the U.S. Postal Inspection Service and the U.S. Immigration and
Naturalization Service (INS).
The U.S. Secret Service, a member of the Northern Virginia
Regional Identification System with a database of over 200,000
fingerprints from 10 Virginia police departments, uses AFIS as
an investigative tool in identifying suspects, for example, by
matching latent prints found on threat letters addressed to the
President.
INS is developing a database of fingerprints for all aliens
intercepted when illegally entering the United States over the
border with Mexico. INS takes a live scan of both index
fingers, photographs the individual, and records certain
biographical data on each subject. In January 1995, INS had
about 75,000 prints in the system and conducted about 2,000
searches a day. According to INS officials, the combined false
positive and error rate\13 is less than 2 percent. The INS
system, whose development began in 1990, is used to identify
aliens that are recidivists, reduce the time required to process
the aliens before returning them to Mexico, and determine
whether an alien is wanted on criminal charges.
--------------------
\10 Latent fingerprints are generally obtained at crime scenes or
from documents or material related to the crimes. These prints
usually occur as isolated finger impressions or as fragmentary parts
of two or three adjacent fingers and are often of poor quality.
\11 Now named the National Institute for Standards and Technology.
\12 A cold search is one conducted with no identifiers available on
the subject/suspect that would assist in narrowing the search within
a system.
\13 A false positive occurs if the system identifies someone that is
not in the database. An error occurs when the system provides a
negative response even though the prints are in the database.
FBI DEVELOPING THE
INTEGRATED AUTOMATED
FINGERPRINT IDENTIFICATION
SYSTEM (IAFIS)
------------------------------------------------------- Appendix I:1.3
The FBI has about 72 million individual criminal and civil 10-print
cards on file and processes 35,000 to 50,000 new 10-print cards each
day. According to an official, the FBI is scanning the inked
hardcopy 10-print cards into the Integrated Automated Fingerprint
Identification System (IAFIS) and digitizing the prints. An FBI
official said that by searching the prints electronically, the FBI
can respond to a routine law enforcement query in about 16 hours and
to special requests in about 2 hours. The system, by searching
digital prints electronically, eliminates the need for the old manual
system and speeds up the process. The FBI plans to completely
automate its system by 1998.
BENEFITS OF FINGERPRINT
TECHNOLOGY FOR THE PRIVATE
SECTOR
------------------------------------------------------- Appendix I:1.4
With the development of live scan equipment and the technology
enabling the electronic storage and transmission of fingerprints, the
private sector has started to use fingerprint technology. Figure I.3
shows one example of a live scan fingerprint reader.
Figure I.3: Live Scan
Fingerprint Reader
(See figure in printed
edition.)
For example, one company that we visited is planning to use
fingerprint technology to verify an individual's identity in
self-check grocery stores, eliminating the need for checkout clerks.
According to company officials, they also plan to institute a store
credit card. The card will include a digitally encoded fingerprint
on the magnetic stripe that can be used with a live scan fingerprint
device to verify ownership of the card prior to charging the
purchases on the card. The system will verify the print encoded on
the magnetic stripe with the customer's fingerprint.
Officials from the firm marketing this technology told us that the
digitally encoded fingerprint uses only 100 bytes of space on a
magnetic stripe. This is particularly important, they noted, because
the stripe has limited storage capacity due to the credit/banking
information on the card. The officials also said that the card would
not be usable if anyone attempts to encode a new print on the card.
In addition, after encoding the fingerprint, the card can be
presented at any location that has installed the company's
fingerprint reader. According to officials, the company is marketing
the technology for such uses as employee time cards, alarm systems,
entry doors, computer files, credit or electronic benefit transfer
cards, checks, driver licenses, social security cards, and
controlled-substance (cigarettes and liquor) vending machines.
HAND GEOMETRY
--------------------------------------------------------- Appendix I:2
Hand geometry is based on the premise that each individual's hands,
although changing over time, remain characteristically the same. An
electronic hand geometry device being used at selected locations by
the Federal Bureau of Prisons stores a template of the hand in the
device's memory. The hand geometry unit measures the height of the
hand, the distance between knuckles, and other information that is
converted to an algorithm. According to agency officials, the more
times a hand geometry device reads or scans a particular hand the
more accurate the reading on that hand becomes. The hand geometry
unit will reject any verification attempts with a hand that has not
been stored in memory. Figure I.4 shows a hand geometry device.
Figure I.4: Hand Geometry
Device
(See figure in printed
edition.)
The Bureau of Prisons is using a personal computer on a local area
network to capture the hand geometry, photograph, and biographical
data of each staff member, visitor, and inmate. The agency provides
a photo identification card with a magnetic stripe encoded with a PIN
to each individual that is in the system. In this way, a prison
checkpoint person can use facial recognition and a database match by
having the card read electronically during the hand geometry
verification process. According to agency officials, their system of
eight hand geometry units, cameras, and necessary software costs
about $90,000, exclusive of the costs for the personal computers or
local area network.
In addition, a Bureau of Prisons official told us that the system has
been very reliable. For example, the agency had conducted over
200,000 hand geometry checks at one prison site with no errors. As
currently designed, if the system does not match an individual's hand
after three attempts, a sound alerts the guards, a "reject" appears
on the computer screen, and the photograph and file of the
appropriate individual appear on screen. The Bureau of Prisons plans
to expand the hand geometry program to its Washington, D.C.,
headquarters and later to establish a wide area network allowing
headquarters staff immediate access to the database that will provide
timely information on the location of inmates and correctional staff
in the event of a riot or disturbance at a federal correctional
institution.
RETINA SCAN
--------------------------------------------------------- Appendix I:3
Retina scan is being used for both access control and for identifying
and releasing felons from custody. Retina identification is based on
a medical finding in 1935 that no two persons have the same pattern
of blood vessels in their retinas. The retina scan device was
developed by an ophthalmologist and is used to capture the unique
pattern of blood vessels in a person's eye. The data are converted
to an algorithm and then stored in a computer or in the scanner's
memory. Enrollment with a retina scan device can be done with one or
both eyes depending upon the user's requirements. For identity
verification, an individual would enter a PIN and place his or her
eye over the lens in proper alignment for scanning. The reading is
compared with the eye signature stored with the PIN in the system.
If there is a match, the individual is identified. The system scores
the eye signature in percentages from 0 to 100 with a minimum score
of 70 percent being recommended by the manufacturer for an accurate
eye measurement. According to a local police official familiar with
retina scan, an eye signature pattern is adversely affected only by a
serious eye illness or injury, such as a detached retina, or eye
surgery. Figure I.5 shows retina scan devices.
Figure I.5: Retina Scan
Devices
(See figure in printed
edition.)
(See figure in printed
edition.)
In 1990, a sheriff's office in a large urban, Midwestern county
purchased and installed retina scanners in its jurisdiction for the
purpose of prisoner identification and release. At that time, the
cost of the system was about $500,000 and included 23 scanners and
other necessary hardware to run an integrated system. An individual
scanner was priced at about $7,000, but the manufacturer's current
model is about $3,500.
According to an official at the sheriff's office, approximately 250
to 300 prisoners are scanned per day, and the database includes more
than 300,000 eye signature templates. A system search on an eye
template takes about 2 minutes. However, a response time of several
seconds was evident in a retina scan device being used for access
control by legal staff in the same jurisdiction. The device being
used by the legal staff was not connected to a large database but
used its stand-alone memory capacity and contained a very limited
number of templates.
Although the manufacturer has recommended requirements for the
scanner's maintenance and environmental operating conditions related
to temperature and humidity, a sheriff's office official indicated
that the scanners were not cleaned on a regular basis nor were
measurements on temperature or humidity taken in the 5 years since
the equipment had been installed. The official advised that the
equipment had already exceeded the manufacturer's duty life and if a
scanner is not operating, it is usually due to a dirty lens or a
power surge. The sheriff's office representative indicated that a
proposal had been submitted to the county to upgrade the system's
equipment, including the installation of about 210 new retina scan
units.
VOICE VERIFICATION
--------------------------------------------------------- Appendix I:4
Voice verification is primarily used to secure building access.
According to one vendor, the voice verification process utilizes such
characteristics of the voice as bass and treble tones, vibration in
the larynx, throat and nasal tones, and air pressure of the person
speaking. According to the vendor, the premise that these
characteristics are distinct for individual voices is based on
research conducted by another firm about 10 years ago. In addition,
the vendor said that each individual voice print is encoded with a
proprietary algorithm.
One Eastern city we visited used voice verification for employee
access and theft prevention at one of its maintenance buildings. The
system at this facility controls both entry to and exit from the
building and secures the use of an elevator in the same building. An
employee gains access by using a dedicated phone at each door. After
the phone is taken off the hook, the voice verification system
prompts the individual to enter his or her PIN number. When the
system locates the employee's voice record, it prompts the individual
to recite a password; and if verified, the system automatically
unlocks the door. After three unsuccessful attempts, the system
responds with "unable to verify." Access rejection usually results
from the use of an incorrect PIN number or password. The system
automatically records both failed and successful attempts by date and
time. Since the system must be accessed for entry or exit, it can be
used for employee time and attendance. The system is used to control
access by particular groups for specific time periods; for example,
the janitorial crew's access is limited to the time they report to
work in the evenings.
According to the city official with whom we spoke, three voice reads
were entered by each enrollee to create a voice template for
verification purposes. He also advised that the system has an
adjustable threshold for verification. The official attributes no
voice access problems to individual voice changes due to such things
as colds or sinus problems. The same official told us that since the
system's installation in 1993, it has required very little
maintenance other than the replacement of keypads on the phones used
with the system. He indicated that the city paid less than the
market value for its system, but a standard voice verification access
system that includes a personal computer, software, and other
hardware for controlling door locks is priced at about $50,000. But,
he added, the system has saved the city about $100,000 annually in
security guard costs; and the city plans to add the same system to
another building.
SIGNATURE VERIFICATION
--------------------------------------------------------- Appendix I:5
Electronic signature verification as an identity authentication
technique is being piloted for such applications as access to safe
deposit boxes, check verification, and computer access for purchase
order authorization control. According to one vendor, signature
verification examines the way a signature is written, rather than the
way it looks. The basic premise of the system is that people have a
tendency to write or sign their names in a consistent and unique
manner.
The same vendor told us that the firm's signature verification device
measures four components: (1) signature shape or form, (2) writing
velocity, (3) pen pressure, and (4) the tracking of the pen when it
is lifted off the tablet. These measurements are automatically
encrypted to ensure that an individual's signature cannot be
reproduced. Signature enrollment by an applicant consists of writing
multiple signatures using a pressure sensitive pen on an electronic
pad or tablet to create a template for signature verification. The
signature files or templates can be stored on barcodes and on smart
cards. Normally a four-digit PIN number is assigned to an
individual's signature template for quick retrieval and comparison;
and the process takes no longer than a normal ATM transaction. The
signature verification system compares a signature written on the
electronic tablet with the signature template previously created by
the user, which is stored in a personal computer or within an
integrated system. Figure I.6 displays an example of a signature
verification device.
Figure I.6: Signature
Verification Device
(See figure in printed
edition.)
According to one vendor with whom we spoke, the signature
verification system "learns" the changes or variations that occur in
an individual's signature over time and adapts to those changes. In
addition, the vendor said that the system can record and store
signatures in any style or language. Experimentation with the
writing of Xs and scrawls by illiterate persons and subsequent
verification appears to work, according to one of the vendors. The
same vendor says the system has an adjustable acceptance range for
allowing leeway for signature acceptance.
During our observations of two different signature verification
systems, one user had repeated difficulty during the verification
process after enrollment. According to one vendor, approximately 1
percent of the population has difficulty with signature enrollment
and verification; and in such cases, additional templates may have to
be taken and the system acceptance range adjusted for easier
acceptance. On the basis of pilot tests and sites where the system
has been applied, the same vendor reports a 0.7-percent false
rejection rate and a 0.4-percent false positive rate.
A research and development firm reports successful results with a
prototype signature verification system. The firm enrolled 4,000
users in the project and found that the system accepted 98 percent of
all sign-on cases (about 300 a day) and 94 percent required only one
signature attempt. According to the firm's report, the project data
indicates that the false rejection rate is about 0.1 percent,
including problems with the system algorithm.
(See figure in printed edition.)Appendix II
COMMENTS FROM THE U.S. DEPARTMENT
OF THE TREASURY
=========================================================== Appendix I
(See figure in printed edition.)
MAJOR CONTRIBUTORS TO THIS REPORT
========================================================= Appendix III
OFFICE OF SPECIAL INVESTIGATIONS,
WASHINGTON, D.C.
Houston R. Fuller, Assistant Director
Thomas L. Sipes, Senior Special Agent
M. Jane Hunt, Senior Communications Analyst
OFFICE OF THE GENERAL COUNSEL,
WASHINGTON, D.C.
Leslie J. Krasner, Attorney Adviser
|
