[House Hearing, 116 Congress]
[From the U.S. Government Publishing Office]
ABOUT FACE: EXAMINING THE DEPARTMENT OF HOMELAND SECURITY'S USE OF
FACIAL RECOGNITION AND OTHER BIOMETRIC TECHNOLOGIES, PART II
=======================================================================
HEARING
BEFORE THE
COMMITTEE ON HOMELAND SECURITY
HOUSE OF REPRESENTATIVES
ONE HUNDRED SIXTEENTH CONGRESS
SECOND SESSION
__________
FEBRUARY 6, 2020
__________
Serial No. 116-60
__________
Printed for the use of the Committee on Homeland Security
[GRAPHIC NOT AVAILABLE IN TIFF FORMAT]
Available via the World Wide Web: http://www.govinfo.gov
__________
U.S. GOVERNMENT PUBLISHING OFFICE
41-450 PDF WASHINGTON : 2020
--------------------------------------------------------------------------------------
COMMITTEE ON HOMELAND SECURITY
Bennie G. Thompson, Mississippi, Chairman
Sheila Jackson Lee, Texas Mike Rogers, Alabama
James R. Langevin, Rhode Island Peter T. King, New York
Cedric L. Richmond, Louisiana Michael T. McCaul, Texas
Donald M. Payne, Jr., New Jersey John Katko, New York
Kathleen M. Rice, New York Mark Walker, North Carolina
J. Luis Correa, California Clay Higgins, Louisiana
Xochitl Torres Small, New Mexico Debbie Lesko, Arizona
Max Rose, New York Mark Green, Tennessee
Lauren Underwood, Illinois John Joyce, Pennsylvania
Elissa Slotkin, Michigan Dan Crenshaw, Texas
Emanuel Cleaver, Missouri Michael Guest, Mississippi
Al Green, Texas Dan Bishop, North Carolina
Yvette D. Clarke, New York Jefferson Van Drew, New Jersey
Dina Titus, Nevada
Bonnie Watson Coleman, New Jersey
Nanette Diaz Barragan, California
Val Butler Demings, Florida
Hope Goins, Staff Director
Chris Vieson, Minority Staff Director
C O N T E N T S
----------
Page
Statements
The Honorable Bennie G. Thompson, a Representative in Congress
From the State of Mississippi, and Chairman, Committee on
Homeland Security:
Oral Statement................................................. 1
Prepared Statement............................................. 2
The Honorable Mike Rogers, a Representative in Congress From the
State of Alabama, and Ranking Member, Committee on Homeland
Security:
Oral Statement................................................. 29
Prepared Statement............................................. 30
The Honorable Sheila Jackson Lee, a Representative in Congress
From the State of Texas:
Prepared Statement............................................. 3
Witnesses
Mr. John P. Wagner, Deputy Executive Assistant Commissioner,
Office of Field Operations, U.S. Customs and Border Protection,
U.S. Department of Homeland Security:
Oral Statement................................................. 6
Prepared Statement............................................. 8
Mr. Peter E. Mina, Deputy Officer for Programs and Compliance,
Office of Civil Rights and Civil Liberties, U.S. Department of
Homeland Security:
Oral Statement................................................. 15
Prepared Statement............................................. 17
Dr. Charles H. Romine, Ph.D., Director of the Information
Technology Laboratory, National Institute of Standards and
Technology:
Oral Statement................................................. 20
Prepared Statement............................................. 22
For the Record
The Honorable Bennie G. Thompson, a Representative in Congress
From the State of Mississippi, and Chairman, Committee on
Homeland Security:
Letter From the Electronic Privacy Information Center.......... 76
News Release, U.S. Travel Association.......................... 81
The Honorable Mike Rogers, a Representative in Congress From the
State of Alabama, and Ranking Member, Committee on Homeland
Security:
White Paper by Security Industry Association................... 59
Article From the Information Technology and Innovation
Foundation (ITIF)............................................ 63
The Honorable Sheila Jackson Lee, a Representative in Congress
From the State of Texas:
Article........................................................ 74
The Honorable John Katko, a Representative in Congress From the
State of New York:
Letter From Chad F. Wolf, Acting Secretary, U.S. Department of
Homeland Security............................................ 39
ABOUT FACE: EXAMINING THE DEPARTMENT OF HOMELAND SECURITY'S USE OF
FACIAL RECOGNITION AND OTHER BIOMETRIC TECHNOLOGIES, PART II
----------
Thursday, February 6, 2020
U.S. House of Representatives,
Committee on Homeland Security,
Washington, DC.
The committee met, pursuant to notice, at 10:05 a.m., in
room 310, Cannon House Office Building, Hon. Bennie G. Thompson
[Chairman of the committee], presiding.
Present: Representatives Thompson, Jackson Lee, Langevin,
Payne, Rice, Correa, Small, Rose, Underwood, Slotkin, Green of
Texas, Clarke, Titus, Coleman, Barragan; Rogers, McCaul, Katko,
Walker, Higgins, Lesko, Green of Tennessee, Joyce, and Shaw.
Chairman Thompson. The Committee on Homeland Security will
come to order.
Let me say at the outset a number of our Members are still
en route from the Prayer Breakfast this morning, and they will
join us accordingly, the Ranking Member being one of them.
The committee is meeting today to receive testimony on the
Department of Homeland Security's use of facial recognition and
other biometric technologies.
Without objection, the Chair is authorized to declare the
committee in recess at any point.
Good morning. The committee is meeting today to continue
examining the Department of Homeland Security's use of facial
recognition technology.
The committee held Part I of this hearing in July of last
year, after news that the Department was expanding its use of
facial recognition for varying purposes, such as confirming the
identity of travelers, including U.S. citizens.
As facial recognition technology has advanced, it has
become the chosen form of biometric technology used by the
Government and industry.
I want to reiterate that I am not wholly opposed to the use
of facial recognition technology, as I recognize that it can be
a valuable tool to the homeland security and serve as a
facilitation tool for the Department's various missions.
But I remain deeply concerned about privacy, transparency,
data security, and accuracy of this technology, and want to
ensure those concerns are addressed before the Department
deploys it any further.
Last July, I, along with other Members of this committee,
shared these concerns at our hearings and left this room with
more questions than answers.
In December 2019, the National Institute for Standards and
Technology published a report that confirmed age, gender, and
racial bias in facial recognition algorithms.
NIST, for example, found that depending on the algorithm,
African American and Asian American faces were misidentified 10
to 100 times more than white faces.
Although CBP touts that the match rate for this facial
recognition system is over 98 percent, it is my understanding
that NIST did not test CBP's current algorithm for its December
2019 report.
Moreover, CBP's figures do not account for images of
travelers who could not be captured due to a variety of
factors, such as lighting or skin tone, actually making the
actual match rate significantly lower.
These findings continue to suggest that some of this
technology is not really ready for prime time and requires
further testing before wide-spread deployment.
Misidentifying even a relatively small percentage of the
traveling public could affect thousands of passengers annually
and likely would have a disproportionate effect on certain
individuals. This is unacceptable.
Data security also remains an important concern. Last year
a CBP contractor experienced a significant data breach, which
included traveler images being stolen.
We look forward to hearing more about these lessons CBP
learned from this incident and the steps that it takes to
ensure that biometric data is kept safe.
Transparency continues to be key. The American people
deserve to know how the Department is collecting facial
recognition data and whether the Department is, in fact,
safeguarding their rights when deploying such technology.
That is why we are here 7 months later to continue our
oversight. I am pleased that we again have witnesses from CBP
and NIST before us to provide us with an update and answer our
questions.
We will also have testimony from DHS's Office of Civil
Rights and Civil Liberties. This office is charged with
ensuring the protection of our civil rights and civil liberties
as it relates to the Department's activities, no easy task,
especially these days.
Be assured that under my leadership this committee will
continue to hold the Department accountable for treating all
Americans equitably and ensuring that our rights are protected.
I look forward to a robust discussion with all of the
witnesses, and I thank the Members for joining us today.
[The statement of Chairman Thompson follows:]
Statement of Chairman Bennie G. Thompson
February 6, 2020
The Committee on Homeland Security is meeting today to continue
examining the Department of Homeland Security's use of facial
recognition technology. The committee held Part I of this hearing in
July of last year--after news that the Department was expanding its use
of facial recognition for varying purposes, such as confirming the
identities of travelers, including U.S. citizens. As facial recognition
technology has advanced, it has become the chosen form of biometric
technology used by the Government and industry. I want to reiterate
that I am not wholly opposed to the use of facial recognition
technology, as I recognize that it can be valuable to homeland security
and serve as a facilitation tool for the Department's varying missions.
But I remain deeply concerned about privacy, transparency, data
security, and the accuracy of this technology and want to ensure these
concerns are addressed before the Department deploys it further.
Last July, I--along with other Members of this committee--shared
these concerns at our hearing and left this room with more questions
than answers. In December 2019, the National Institute for Standards
and Technology (NIST) published a report that confirmed age, gender,
and racial bias in some facial recognition algorithms. NIST, for
example, found that depending on the algorithm, African-American and
Asian-American faces were misidentified 10 to 100 times more than white
faces. Although CBP touts that the match rate for its facial
recognition systems is over 98 percent, it is my understanding that
NIST did not test CBP's current algorithm for its December 2019 report.
Moreover, CBP's figure does not account for images of travelers who
could not be captured due a variety of factors such as lighting or skin
tone--likely making the actual match rate significantly lower. These
findings continue to suggest that some of this technology is not ready
for ``prime time'' and requires further testing before wide-spread
deployment.
Misidentifying even a relatively small percentage of the traveling
public could affect thousands of passengers annually, and likely would
have a disproportionate effect on certain individuals. This is
unacceptable. Data security also remains an important concern. Last
year, a CBP subcontractor experienced a significant data breach, which
included traveler images being stolen. We look forward to hearing more
about the lessons CBP learned from this incident and the steps that it
has taken to ensure that biometric data is kept safe. Transparency
continues to be key. The American people deserve to know how the
Department is collecting facial recognition data, and whether the
Department is in fact safeguarding their rights when deploying such
technology. That is why we are here 7 months later to continue our
oversight.
I am pleased that we again have witnesses from CBP and NIST before
us to provide us with an update and answer our questions. We will also
have testimony from DHS's Office for Civil Rights and Civil Liberties.
This office is charged with ensuring the protection of our civil rights
and civil liberties as it relates to the Department's activities--no
easy task, especially these days. Be assured that under my leadership,
this committee will continue to hold the Department accountable for
treating all Americans equitably and ensuring that our rights are
protected.
Chairman Thompson. Other Members are reminded that
statements may be submitted for the record.
[The statement of Honorable Jackson Lee follows:]
Statement of Honorable Sheila Jackson Lee
February 6, 2020
Thank you, Chairman Thompson and Ranking Member Rogers for holding
today's important hearing on ``About Face: Examining the Department of
Homeland Security's Use of Facial Recognition and Other Biometric
Technologies, Part II.''
I look forward to hearing from today's Government witnesses on
DHS's use of facial recognition and other biometric technologies.
Good morning and welcome to our witnesses:
Mr. John Wagner, deputy executive assistant commissioner,
Office of Field Operations, U.S. Customs and Border Protection
(CBP), Department of Homeland Security (DHS),
Mr. Peter Mina, deputy officer for programs and compliance,
Office for Civil Rights and Civil Liberties (CRCL), DHS,
Dr. Charles H. Romine, director, Information Technology
Laboratory, National Institute of Standards and Technology
(NIST), Department of Commerce.
The hearing today provides an opportunity for Members of this
committee to examine DHS's use of biometric technologies, including
facial recognition technology, for Government purposes.
Biometrics is the technical term for body measurements and
calculations.
It refers to metrics related to human characteristics such as
fingerprints, eyes, voice, or other unique features associated with
people.
Biometrics authentication is used in computer science as a form of
identification and access control.
Facial recognition is one of the most popular biometrics.
Facial recognition systems are computer-based security systems,
which are deployed to automatically detect and identify human faces.
Several DHS components have begun expanding their use of facial
recognition technology for purposes ranging from identifying travelers
to general surveillance.
My admiration and respect for the men and women of DHS as public
servants who are our Nation's first line of defense against terrorism
that targets our Nation is well-known.
Securing our Nation's transportation systems, critical
infrastructure, and civil government agencies from cyber threats
requires efficiency and effectiveness of all aspects of recruitment,
training, and retention of professionals.
In the last decade, domestic terrorism has become an increasing
concern in the United States, and these persons are in the United
States, and not coming from overseas.
So there needs to be concern when people of color are targets of
those seeking to do violence to people living within our own Nation's
borders.
In 2018, domestic extremists killed at least 50 people in the
United States, a sharp increase from the 37 extremist-related murders
documented in 2017, though still lower than the totals for 2015 (70)
and 2016 (72).
The 50 deaths made 2018 the fourth-deadliest year on record for
domestic extremist-related killings since 1970.
According to an analysis by the Washington Post, between 2010 and
2017, right-wing terrorists committed a third of all acts of domestic
terrorism in the United States (92 out of 263), more than Islamist
terrorists (38 out of 263) and left-wing terrorists (34 out of 263) put
together.
Recent unpublished FBI data leaked to the Washington Post in early
March 2019 reveal that there were more domestic terrorism-related
arrests than international terrorism-related arrests in both fiscal
year 2017 and fiscal year 2018.
From 2009 to 2018 there were 427 extremist-related killings in the
United States; of those, 73.3 percent were committed by right-wing
extremists, 23.4 percent by Islamist extremists, and 3.2 percent by
left-wing extremists.
In short, 3 out of 4 killings committed by right-wing extremists in
the United States were committed by white supremacists (313 from 2009
to 2018).
The culmination of the 2018 mid-term election was consumed by bombs
placed in the mail addressed to Democrats.
The risks posed by terrorism must be weighed aganist the privacy
and civil liberties concerns raised by the deployment and use of
biometric idetntification systems including facial recognition.
Today, DHS components including TSA, CBP, and ICE interact more
intimately with broad swaths of the public than any other Government
agency, screening over 2 million passengers every day.
On July 7, 2019, the New York Times reported that ICE has been
mining State driver's license records for immigration purposes.
According to this article at least 3 States that offer driver's
licenses to undocumented immigrants, ICE officials have requested to
comb through State repositories of license photos, according to newly-
released documents.
At least 2 of those States, Utah and Vermont, complied, searching
their photos for matches, those records show.
In the third State, Washington, agents authorized administrative
subpoenas of the Department of Licensing to conduct a facial
recognition scan of all photos of license applicants, though it was
unclear whether the State carried out the searches.
In Vermont, agents only had to file a paper request that was later
approved by Department of Motor Vehicles employees.
Over 50 percent of all Americans are included in State Department
of Motor Vehicle records.
Members of this Committee understand that several components within
the Department gather and collect biometric information.
DHS uses biometrics for the purposes of identity verification, and
it has looked to increase its use of technologies for such purposes.
Currently, TSA front-line workers and airline employees manually
compare the traveler in front of them to the photo identification
provided.
TSA seeks to leverage facial recognition technology to automate the
identity verification process to enhance security effectiveness,
improve operational efficiency, and streamline the traveler experience.
TSA has demonstrated an interest in using facial recognition to
validate the identity of TSA PreCheck passengers who have voluntarily
provided biometric information to TSA.
TSA has begun capturing photographs of passengers enrolling or
renewing enrollments in PreCheck.
The U.S. Citizenship and Immigration Services (USCIS) has long
collected fingerprints and pictures of applicants for immigration
benefits.
CBP has begun implementing a biometric entry-exit system that
relies on facial recognition for verifying a traveler's identification,
including U.S. citizens.
TSA is interested in using facial recognition to validate the
identity of TSA PreCheck passengers who have voluntarily provided
biometric information to TSA and, eventually, to verify passenger
identity for standard screening, including for domestic travel.
Beyond confirming an individual's identity, some components have
been using, or are contemplating the use of, facial recognition
technology to surveil a crowd of people for law enforcement purposes.
Since November, the United States Secret Service has been
conducting a facial recognition pilot on a limited basis at the White
House to search the faces of individuals visiting the complex or
passing by on public streets and parks.
Emails from Immigration and Customs Enforcement (ICE) officials
became public, which detailed meetings with Amazon over its facial
recognition platform ``Rekognition'' and its possible use on the
Southern Border.
In 2018, this same system was reported to have falsely identified
28 Members of Congress as having a match to known criminals.
The committee must fully understand the limitations of facial
recognition systems.
Although algorithms may be well-developed and work extremely well,
if the technology is applied to data that is of poor quality or have
weak technical standards then the output can be worthless.
If the underlying technology is not the right match for the
intended purpose of facial recognition or discernment then the system
will fail.
The National Institute of Science (NIST) has done admirable work in
producing 3 reports on the topic of facial recognition, and I look
forward to learning more about their work in scoring their performance.
I gather from their efforts that they provide a technical
assessment of the facial recognition applications brought to them for
analysis by other Federal Government agencies.
NIST does not see the algorithms against their own data set and
observe the outcomes to assess the performance of facial recognition
applications.
Additionally, agency can request facial recognition for any
vendor--the only requirement is that the agency wait 3 months before
making a second request.
Use of facial recognition technology is expanding within and
outside of the Government, which raises concerns about privacy, civil
liberties, and accuracy in the application of Federal administrative
procedures that may affect a range of agency decision making such as
the right to travel--and extend into applications used to determine
qualifications for Federal benefits programs such as Social Security,
Medicare/Medicaid, or Veterans programs.
These concerns relate to the accuracy, reliability, and fairness to
those who may be subject to Federal use of facial recognition systems
are not trivial.
Collection and storage of facial images can occur with or without
the consent of data subjects.
There is no law governing facial image capture for Government
purposes.
Biometric facial recognition systems deployed at public gatherings
can be used to support facial image capture for storage and later use
without the knowledge or permission of data subjects.
The ``one to many'' application of facial recognition technology
involves--taking one image of a person and comparing it to stored
images of perhaps hundreds or thousands of people to successfully
identify a person is the ``Holy Grail'' of facial recognition.
There are law enforcement, National security, defense, and homeland
security applications that would benefit from the success of accurately
identifying individuals.
There are also commercial applications for being able to with a
high degree of accuracy pick a face out of a crowd.
Because there is such strong interest in solving the problems of
face recognition the Congress does need to keep track of developments
in this area.
The systems of facial recognition currently available have flaws
and are not as accurate or reliable as they might become as the
technology evolves.
Today, we need laws that govern how Federal biometric systems can
be deployed and reign in how the data collected might be used.
The committee needs to know where DHS is getting the images it is
using and whether third-party vendors allow the agency to avoid Privacy
Act considerations.
It is incumbent upon our committee to provide the necessary
guidance to DHS on how these technologies can be used when
Constitutionally-protected activities are involved.
DHS components have proceeded with the acquisition and deployment
of facial recognition technology with little guidance or oversight from
the Congress or other Federal entities.
The topic of today's hearing is important and I thank the Chairman
for his foresight in bringing today's witnesses before the committee.
I look forward to the testimony of today's witnesses.
Thank you.
Chairman Thompson. I welcome our panel of witnesses. Our
first witness, Mr. John Wagner, currently serves as the deputy
executive assistant commissioner for the Office of Field
Operations, U.S. Customs and Border Protection. In his current
role, he oversees nearly 30,000 Federal employees and manages
programs related to immigration, customs, and commercial trade-
related CBP missions.
Mr. Peter Mina is a deputy officer for programs and
compliance at the Office of Civil Rights and Civil Liberties.
Mr. Mina previously served as chief of the Labor and Employment
Law Division for U.S. Immigration and Customs Enforcement.
Dr. Charles Romine is the director of the Information
Technology Laboratory at the National Institute of Standards
and Technology. In this position, he oversees a research
program that focuses on testing and interoperability, security,
usability, and reliability of information systems.
Without objection, the witnesses' full statements will be
inserted in the record.
I now ask each witness to summarize his statement for 5
minutes, beginning with Mr. Wagner.
STATEMENT OF JOHN P. WAGNER, DEPUTY EXECUTIVE ASSISTANT
COMMISSIONER, OFFICE OF FIELD OPERATIONS, U.S. CUSTOMS AND
BORDER PROTECTION, U.S. DEPARTMENT OF HOMELAND SECURITY
Mr. Wagner. Good morning. Chairman Thompson, Ranking Member
Rogers, Members of the committee, thank you for the opportunity
to testify here before you today on behalf of U.S. Customs and
Border Protection.
I am looking forward to the opportunity to discuss the
recent NIST report with you today. Since CBP is using an
algorithm from one of the highest-performing vendors identified
in the report, we are confident that our results are
corroborated with the findings of this report.
More specifically, the report indicates while there is a
wide range of performance, of the 189 different algorithms that
NIST reviewed, the highest-performing algorithms had minimal to
undetectable levels of demographic-based error rates.
The report also highlights some of the operational
variables that impact error rates, such as gallery size, photo
age, photo quality, numbers of photos of each subject in the
gallery, camera quality, lighting, human behavior factors. All
influence the accuracy of an algorithm.
That is why CBP has carefully constructed the operational
variables in the deployment of the technology to ensure we can
attain the highest levels of match rates, which remain in the
97 to 98 percent range.
One important note is that NIST did not test the specific
CBP operational construct to measure the additional impact
these variables may have, which is why we have recently entered
into an MOU with NIST to evaluate our specific data.
But as we build out the Congressionally-mandated biometric-
based entry/exit system, we are creating a system that not only
meets the security mandate, but also in a way that is cost-
effective, feasible, and facilitative for international
travelers.
Identity requirements are not new when crossing the border
or taking an international flight. Several existing laws and
regulations require travelers to establish their identity and
citizenship when entering and departing the United States.
CBP employs biographic and biometric-based procedures to
inspect the travel documents presented by individuals to verify
the authenticity of the document and determine if it belongs to
the actual person presenting it.
Again, these are not new requirements. The use of facial
comparison technology simply automates the process that is
often done manually today.
The shortcomings of human manual review in making facial
comparisons are well-documented. Humans are prone to fatigue,
sometimes have biases they may not even realize to include own
race and gender biases.
Fingerprint biometrics have also documented gaps in their
performance. There is a small percentage of people that we see
we cannot capture fingerprints from, and there are studies that
document this, as well, as well as demographic correlations,
most notably based on age.
We are all well aware of the issues of common names when we
rely on a biographic-based vetting scheme alone. So no one
system by itself is perfect.
However, since the United States, along with many other
countries, put a digital photograph into the electronic chip on
a passport, it would seem to make prudent sense that the
technology may be useful in determination of the rightful
document holder.
It is more difficult today to forge or alter a legitimate
passport as security features are more stronger than they were
10 or 15 years ago, but we are still vulnerable to a person
using a legitimate document, particularly a U.S. travel
document, that is real but belongs to someone else.
Using facial comparison technology to date we have
identified 252 imposters, to include people using 75 genuine
U.S. travel documents.
The privacy continues to be integral to our biometric
mission. CBP is compliant with the terms of the Privacy Act of
1974, as amended, the E-Government Act of 2002, the Homeland
Security Act of 2002, the Paperwork Reduction Act of 1995, and
departmental policies that govern the collection, use, and
maintenance of personally-identifiable information.
CBP recently published updates to the appendices in the
privacy impact assessment covering this program, and Systems of
Record notices have been published on the databases to process
and store the information.
We have met 3 times with representatives of the privacy
advocacy community, as well as discussions with the Privacy and
Civil Liberties Oversight Board, and the DHS Privacy and
Integrity Advisory Committee.
In November, CBP submitted to the Office of Management and
Budget a rulemaking that would solicit public comments on the
proposed regulatory updates and amendments to the Federal
regulations.
One final note is that our private-sector partners, the
airlines and the airports, must agree to documented specific
CBP business requirements if they are submitting photographs to
CBP as part of this process. These requirements include a
provision that images must be deleted after they are
transmitted to CBP and may not be retained by the private
stakeholder.
After the devastating attacks of September 11, we as a
Nation asked, ``How can we make sure this never happens
again?'' As part of that answer, the 9/11 Commission report
recommended that DHS should complete as quickly as possible a
biometric entry/exit screening system, and that it was, ``an
essential investment in National security.''
CBP is answering that call in carrying out the duties
Congress has given us by continuing to strengthen its biometric
efforts along the travel continuum and verifying that people
are who they say they are.
I thank you for the opportunity to appear today, and I look
forward to your questions.
[The prepared statement of Mr. Wagner follows:]
Prepared Statement of John P. Wagner
February 6, 2020
Chairman Thompson, Ranking Member Rogers, and Members of the
committee, thank you for the opportunity to testify before you on the
efforts of U.S. Customs and Border Protection (CBP) to better secure
our Nation by incorporating biometrics into our comprehensive entry-
exit system, and to identify overstays in support of our border
security mission.
CBP has received public support for its use of biometrics from the
International Air Transit Association, the World Travel and Tourism
Council, and the Department of Commerce Travel and Tourism Advisory
Board.\1\ With international air travel growing at 4.9 percent per year
and expected to double by 2031, and with an increasingly complex threat
posture, CBP must innovate and transform the current travel processes
to handle this expanding volume. Facial comparison technology will
enable CBP and travel industry stakeholders to position the U.S. travel
system as best in class, in turn, driving the continued growth in air
travel volume.
---------------------------------------------------------------------------
\1\ International Air Transport Association, ``Resolution: End-to-
end Seamless Travel across Borders Closer to Reality'' (June 2, 2019).
www.iata.org/en/pressroom/pr/2019-06-02-06/. World Travel & Tourism
Council, ``Gloria Guevara: `We must act and assign priority and
resources to biometrics' ''. March 6, 2019. www.wttc.org/about/media-
centre/press-releases/press-releases/2019/we-must-act-and-assign-
priority-and-resources-to-biometrics/. United States Travel and Tourism
Advisory Board, letter to Commerce Secretary, Wilbur Ross, containing
challenges and recommendation on U.S. Government-private industry
partnerships on biometric technology (April 29, 2019). https://
legacy.trade.gov/ttab/docs/TTAB_Biometrics%20Recommenda-
tions%20Letter_042919.pdf.
---------------------------------------------------------------------------
As authorized in several statutes and regulations, CBP is
Congressionally-mandated to implement a biometric entry-exit system.\2\
Prior to the Consolidated and Further Continuing Appropriations Act of
2013 (Public Law 113-6), which transferred the biometric exit mission
from the Department of Homeland Security's (DHS) United States Visitor
and Immigration Status Indicator Technology (US-VISIT) Program within
the National Protection and Programs Directorate (NPPD) to CBP, the
U.S. Government and the private sector were developing independent
biometrics-based schemes for administering the entry-exit program
responsibilities. These varied and often uncoordinated investments
relied on multiple biometrics and required complicated enrollment
processes. Public and private-sector entities developed separate uses
for biometrics, each with varying privacy risks and accountability
mechanisms. In 2017, CBP developed an integrated approach to the
biometric entry-exit system that other U.S. Government agencies with
security functions, such as TSA, as well as travel industry
stakeholders such as airlines, airports, and cruise lines, could
incorporate into their respective mission space.
---------------------------------------------------------------------------
\2\ Statutes that require DHS to take action to create an
integrated entry-exit system: Sec. 2(a) of the Immigration and
Naturalization Service Data Management Improvement Act of 2000 (DMIA),
P.L. 106-215, 114 Stat. 337; Sec. 110 of the Illegal Immigration Reform
and Immigrant Responsibility Act of 1996, P.L. 104-208, 110 Stat. 3009-
546; Sec. 205 of the Visa Waiver Permanent Program Act of 2000, P.L.
106-396, 114 Stat. 1637, 1641; Sec. 414 of the Uniting and
Strengthening America by Providing Appropriate Tools Required to
Intercept and Obstruct Terrorism Act of 2001 (USA PATRIOT Act), P.L.
107-56, 115 Stat. 272, 353; Sec. 302 of the Enhanced Border Security
and Visa Entry Reform Act of 2002 (Border Security Act), P.L. 107-173,
116 Stat. 543, 552; Sec. 7208 of the Intelligence Reform and Terrorism
Prevention Act of 2004 (IRTPA), P.L. 108-458, 118 Stat. 3638, 3817;
Sec.711 of the Implementing Recommendations of the 9/11 Commission Act
of 2007, P.L. 110-53, 121 Stat. 266, 338; and Sect. 802 of the Trade
Facilitation and Trade Enforcement Act of 2015, P.L. 114-125, 130 Stat.
122, 199. In addition, through the Consolidated Appropriations Act of
2016 and the Bipartisan Budget Act of 2018, Congress authorized up to
$1 billion in visa fee surcharges through 2027 to support biometric
entry/exit. P.L. 114-113 129 Stat. 2242 (December 17, 2015); P.L. 115-
123 132 Stat. 64 (February 9, 2018).
---------------------------------------------------------------------------
CBP offered relevant stakeholders an ``identity as a service''
solution that uses facial comparison to automate manual identity
verification, thereby harmonizing the data collection and privacy
standards each stakeholder must follow. This comprehensive facial
comparison service leverages biographic and biometric data, both of
which are key to support CBP's mission, to fulfill the Congressional
biometric entry-exit mandate while using the system to support air
travel, improve efficiency, and increase the efficacy of identity
verification. CBP has been testing options to leverage biometrics at
entry and departure, specifically through the use of facial comparison
technology.\3\ These technologies enhance the manual process used today
by making it more efficient, accurate, and secure. Using data that
travelers are already required by statute to provide, the automated
identity verification process uses facial comparison to identify those
who are traveling on falsified or fraudulent documents as well as those
seeking to evade screening. These are the individuals who present
public safety or National security threats or have overstayed their
authorized period of admission.
---------------------------------------------------------------------------
\3\ DHS/CBP (November 2018), DHS/CBP/PIA-056 Traveler Verification
Service. (945.31 KB).
---------------------------------------------------------------------------
previous efforts to launch a biometric exit system
Prior to the Consolidated and Further Continuing Appropriations Act
of 2013 (Public Law 113-6), which transferred the biometric exit
mission from DHS headquarters to CBP, the U.S. Government and the
private sector were already developing independent biometric solutions
for administering entry-exit programs. For example, from January 2004
through May 2007, DHS placed kiosks between security checkpoints and
airline gates to collect travelers' fingerprint biometrics. The
traveler had the responsibility to find and use the devices, while
airports where the kiosks were deployed provided varying degrees of
support. In 2008, DHS issued a Notice of Proposed Rulemaking (NPRM)
that proposes commercial air and vessel carriers collect biometric
information from certain aliens departing the United States and submit
this information to DHS within a certain time frame. Most comments
opposed the adoption of the proposed rule, citing cost and feasibility.
Among other comments was the suggestion that biometrics collection
should strictly be a Governmental function. The suggestion was made
that the highly competitive air industry could not support a major new
process of biometric collection on behalf of the Government, and that
requiring air carriers to collect biometrics was not feasible and would
unfairly burden air carriers and airports. Additionally, as directed by
Congress, from May through June 2009, DHS operated 2 biometric exit
pilot programs in which CBP used a mobile device to collect biometric
exit data at departure gates while TSA collected it at security
checkpoints.
DHS concluded from the NPRM comments and pilot programs that it was
generally inefficient and impractical to introduce entirely new
Government processes into an existing and familiar traveler flow,
particularly in the air environment. DHS also concluded that the use of
mobile devices to capture electronic fingerprints would be extremely
resource-intensive. This information helped frame our concept for a
comprehensive biometric entry-exit system that would avoid adding new
processes; utilize existing infrastructure; leverage existing
stakeholder systems, processes, and business models; leverage passenger
behaviors and expectations; and utilize existing traveler data and
existing Government information technology infrastructure.
cbp's integrated approach to a comprehensive biometric entry-exit
system
Leveraging CBP's current authorities, we are executing
Congressional mandates to create and test an integrated biometric
entry-exit system using facial comparison technology. This technology
uses existing advance passenger information along with photographs
already provided to the Government by international travelers to create
``galleries'' of facial image templates that correspond with the
individuals expected on international flights arriving or departing the
United States. These photographs may be derived from passport
applications, visa applications, or interactions with CBP at a prior
border inspection.\4\ Once the gallery is created based on the advance
information, the biometric comparison technology compares a template of
a live photograph of the traveler--taken where there is clear
expectation and authority that a person will need to provide
documentary evidence of their identity--to the gallery of facial image
templates.
---------------------------------------------------------------------------
\4\ Department of State, Consular Consolidated System, ``Privacy
Impact Assessment: Consular Consolidated Database'' (January 29, 2020).
https://2001-2009.state.gov/documents/organization/93772.pdf.
---------------------------------------------------------------------------
For technical demonstrations at the land border, air entry, and
some air exit operations, CBP cameras take photographs of travelers.
These tests have been extended on a voluntary basis to exempt certain
aliens and U.S. citizens.\5\ Participation provides a more accurate and
efficient method to verify identity and citizenship. In other air exit
and seaport demonstrations, CBP does not take the photographs. Instead,
specified partners, such as commercial air carriers, airport
authorities, and cruise lines, take photographs of travelers and
transmit the images to CBP's facial matching service. These partners
use their own camera operators and technology that meets CBP's
technical and security requirements. These tests occur on a voluntary
basis and are consistent with that partner's contractual relationship
with the traveler.
---------------------------------------------------------------------------
\5\ Under Scope of examination, Alien applicants for admission, 8
C.F.R. 235.1(f)(1)(ii) and Requirements for biometric identifiers
from aliens on departure from the United States, 8 C.F.R.
215.8(a)(1), CBP may require certain aliens to provide biometric
identifiers to confirm their admissibility or, at specified airports,
their departure. Some aliens are exempt from the requirement to provide
biometrics. This includes Canadians, under Sect.101(a)(15)(B), who are
not otherwise required to present a visa or be issued a Form I-94 or
Form I-95; aliens younger than 14 or older than 79 on the date of
admission; aliens admitted A-1, A-2, C-3 (except for attendants,
servants, or personal employees of accredited officials), G-1, G-2, G-
3, G-4, NATO-1, NATO-2, NATO-3, NATO-4, NATO-5, or NATO-6 visas; and
certain Taiwan officials and members of their immediate families who
hold E-1 visas, unless the Secretary of State and the Secretary of
Homeland Security jointly determine that a class of such aliens should
be subject to the requirements of paragraph (d)(1)(ii); classes of
aliens to whom the Secretary of Homeland Security and the Secretary of
State jointly determine the requirement shall not apply; or an
individual alien to whom the Secretary of Homeland Security, the
Secretary of State, or the Director of Central Intelligence determines
this requirement shall not apply.
---------------------------------------------------------------------------
Biometric entry-exit is not a surveillance program. CBP does not
use hidden cameras. CBP uses facial comparison technology to ensure a
person is who they say they are--the bearer of the passport they
present. This technology provides a seamless way for in-scope travelers
to meet the requirement to provide biometrics upon departure from the
United States. Travelers are aware their photos are being taken and
that they can opt out as described below. CBP uses facial comparison
technology only where a current identity check already exists. CBP
works closely with partner air carriers and airport authorities to post
privacy notices and provide tear sheets for impacted travelers and
members of the public in close proximity to the cameras and operators,
whether the cameras are owned by CBP or the partners.
The imposter threat--or the use of legitimate documents that do not
belong to the bearer--continues to be a challenge for CBP. U.S.
passports are the most prized version of an imposter document because--
until recently--there was no biometric comparison between the person
presenting the document and the owner of the document. As document
security standards have increased in the past 20 years, it has become
much more difficult to plausibly forge or alter a legitimate document.
As a result, those who wish to evade detection seek to use legitimate
documents that belong to someone else. U.S. citizens are not required
to provide fingerprint biometrics for entry into the country whereas
foreign nationals may be required to do so.
CBP is authorized to require ``in-scope'' aliens to provide
biometric identifiers.\6\ For entry, CBP uses cameras and facial
comparison technology during the inspection process. CBP operates
facial comparison technology pilots at exit in certain land and sea
ports and some airports.\7\ This technology provides the travel
industry with the tools to verify traveler identity and transmit
information to CBP.\8\ We have identified best practices from the prior
DHS work as well as from our international partners and used them in
the biometric exit system design to avoid an inefficient two-step
process that requires multiple biometrics to verify traveler identity.
---------------------------------------------------------------------------
\6\ ``In scope'' aliens may be required to provide biometric
identifiers to confirm their admissibility, or, at specified airports,
their departure in accordance with Inspection of Persons Applying for
Admission, Scope of examination, Alien applicants for admission, 8
C.F.R. 235.1(f)(1)(ii) and Requirements for biometric identifiers
from aliens on departure from the United States, 8 C.F.R.
215.8(a)(1).
\7\ Requirements for biometric identifiers from aliens on departure
from the United States, 8 C.F.R. 215.8(a)(1).
\8\ Numerous statutes require advance electronic transmission of
passenger and crew member manifests for commercial aircraft and
commercial vessels. These mandates include, but are not limited to Sec.
115 of the Aviation and Transportation Security Act (ATSA), P.L. 107-
71, 115 Stat. 597; Passenger manifests, 49 U.S.C. 44909 (applicable
to passenger and crew manifests for flights arriving in the United
States); Sec. 402 of the Enhanced Border Security and Visa Entry Reform
Act of 2002 (EBSVERA), P.L. 107-173, 116 Stat. 543; List of alien and
citizen passengers arriving and departing, 8 U.S.C. 1221; and
Examination of merchandise, 19 U.S.C. 1499.
---------------------------------------------------------------------------
CBP understood the need to build a system that all stakeholders
within the travel continuum could participate in without building their
own independent system--one that could expand to other mission areas
outside of the biometric exit process. To address these challenges and
satisfy the Congressional mandate, we are working closely with our
partners to integrate biometrics with existing identity verification
requirements to the extent feasible.\9\ Facial comparison technology
can match more than 97 percent of travelers through the creation of
facial galleries.\10\ The match rate is based on the percentage of
travelers with a valid encounter photo who were successfully matched to
a gallery photo.\11\
---------------------------------------------------------------------------
\9\ Ibid.
\10\ Department of Homeland Security Fiscal Year 2018 Entry/Exit
Overstay Report, https://www.dhs.gov/sites/default/files/publications/
19_0417_fy18-entry-and-exit-overstay-report.pdf.
\11\ DHS/CBP (November 2018), DHS/CBP/PIA-056 Traveler Verification
Service. (945.31).
---------------------------------------------------------------------------
While CBP's primary responsibility is National security, we must
also facilitate legitimate trade and travel. The use of facial
comparison technology has enabled CBP to not only address a National
security concern head-on by enhancing identity verification but to
simultaneously improve the traveler experience throughout the travel
continuum. CBP engineered a biometric exit solution that gives not only
CBP, but TSA and industry stakeholders such as airlines and airports,
the ability to automate manual identity verification. This may include
departure gates, debarkation (arrival) areas, airport security
checkpoints, and Federal Inspection Services areas.
CBP uses only photos collected from cameras deployed specifically
for this purpose and does not use photos obtained from closed-circuit
television or other live or recorded video. As the facial comparison
technology automates the manual identity verification process in place
today, it allows CBP and its stakeholders to make quicker and more
informed decisions. In August 2019, CBP and TSA provided this committee
a comprehensive report on the program that included material on the
operational and security benefits of the biometric entry-exit system,
CBP and TSA's efforts to address privacy concerns and potential
performance differential errors, and a comprehensive description of
audits performed.\12\
---------------------------------------------------------------------------
\12\ DHS, ``Transportation Security Administration and Customs and
Border Protection: Deployment of Biometric Technologies, Report to
Congress'' (August 30, 2019 www.tsa.gov/sites/default/files/
biometricsreport.pdf.
---------------------------------------------------------------------------
cbp authorities
As described above, numerous Federal statutes require DHS to create
an integrated, automated biometric entry and exit system that records
the arrival and departure of aliens, compares the biometric data to
verify their identities, and authenticates travel documents. Most
recently, in 2017, Executive Order 13780 called for the expedited
completion of the biometric entry-exit data system.\13\ DHS has broad
authority to control alien travel and to inspect aliens under various
provisions of the Immigration and Nationality Act of 1952 (INA), as
amended.\14\ As part of CBP's authority to enforce U.S. immigration
laws, CBP is responsible for interdicting individuals illegally
entering or exiting the United States; facilitating and expediting the
flow of legitimate travelers; and detecting, responding to, and
interdicting terrorists, drug smugglers, human smugglers, traffickers,
and other persons who may undermine the security of the United States
at entry.
---------------------------------------------------------------------------
\13\ Other statues that require DHS to create an integrated entry-
exit system include: Sect. 2(a) of the Immigration and Naturalization
Service Data Management Improvement Act of 2000 (DMIA), P.L. 106-215,
114 Stat. 337; Sec. 205 of the Visa Waiver Permanent Program Act of
2000, P.L. 106-396, 114 Stat. 1637, 1641; and Sec. 414 of the Uniting
and Strengthening America by Providing Appropriate Tools Required to
Intercept and Obstruct Terrorism Act of 2001 (USA PATRIOT Act), P.L.
107-56, 115 Stat. 272, 353.
\14\ Biometric entry and exit data system, 8 U.S.C. 1365b
mandates the creation of an integrated and comprehensive system. The
entry and exit data system shall include a requirement for the
collection of biometric exit data for all categories of individuals
required to provide biometric entry data. As a result, if a certain
category of individuals is required to provide biometrics to DHS on
entry as part of the examination and inspection process, the same
category of individuals must be required to provide biometrics on exit
as well. DHS may require individuals to provide biometrics and other
relevant identifying information upon entry to, or departure from, the
United States. Specifically, DHS may control alien entry and departure
and inspect all travelers under 215(a) and 235 of the INA (8
U.S.C. 1185, 1225). Aliens may be required to provide fingerprints,
photographs, or other biometrics upon arrival in, or departure from,
the United States, and select classes of aliens may be required to
provide information at any time. See, e.g., INA 214, 215(a), 235(a),
262(a), 263(a), 264(c), (8 U.S.C. 1184, 1185(a), 1225(a), 1302(a),
1303(a), 1304(c)); 8 U.S.C. 1365b. Pursuant to 215(a) of the INA (8
U.S.C. 1185(a)), and Executive Order No. 13323 (December 30, 2003)
(69 FR 241), the Secretary of Homeland Security, with the concurrence
of the Secretary of State, has the authority to require aliens to
provide requested biographic information, biometrics, and other
relevant identifying information as they depart the United States.
---------------------------------------------------------------------------
To effectively carry out its responsibilities under the INA for
both arrivals and departures from the United States, CBP must be able
to conclusively determine if a person is a U.S. citizen or national or
an alien by verifying that the person is the true bearer of his or her
travel documentation. CBP is authorized to take and consider evidence
concerning the privilege of any person to enter, reenter, pass through,
or reside in the United States, or concerning any matter material or
relevant to the enforcement or administration of the INA.\15\ A person
claiming U.S. citizenship must establish that fact to the examining
officer's satisfaction and must present a U.S. passport or alternative
documentation.\16\
---------------------------------------------------------------------------
\15\ Powers of immigration officers and employees, 8 U.S.C.
1357(b).
\16\ Under Scope of examination, 8 C.F.R. 235.1(b), it is
generally unlawful for a U.S. citizen to depart or attempt to depart
from the United States without a valid passport. See also Travel
control of citizens and aliens, 8 U.S.C. 1185(b); and Passport
requirement; definitions, 22 C.F.R. 53.1.
---------------------------------------------------------------------------
To further advance the legal framework, CBP is working to propose
and implement regulatory amendments. CBP is working on a biometric
entry/exit regulation, which will only impact foreign nationals. In
November 2019, CBP transmitted its proposed regulation on biometric
entry/exit to the Office of Management and Budget; we are awaiting
clearance. The rule will go through the full rulemaking process, which
includes a public comment period.
nist facial comparison vendor test: december 2019
CBP has partnered with the National Institute of Standards and
Technology (NIST) to explore facial comparison technology capabilities.
NIST used CBP data that was contained in the OBIM data in its
conclusions issued in a recent demographic differential study. The
study supports what CBP has seen in its biometric matching operations--
that when a high-quality facial comparison algorithm is used along with
high-performing cameras, proper lighting and image quality controls,
face-matching technology can be highly accurate. To ensure higher
accuracy rates, as well as efficient traveler processing, CBP compares
traveler photos to a very small gallery of high-quality images that
those travelers already provided to the U.S. Government to obtain a
passport or visa.
CBP uses only one of the 189 face comparison algorithms evaluated
by NIST and produced by NEC Corporation. As the report demonstrates,
NIST confirmed that the NEC algorithm that NIST tested is high-
performing and ranked first or second in most categories evaluated,
including match performance in galleries that are much bigger than
those used by CBP.\17\ The NIST performance metrics described in the
report are consistent with CBP operational performance metrics for
entry-exit. CBP's operational data continues to show there is no
measurable differential performance in matching based on demographic
factors. The NIST report shows a wide range in accuracy across
algorithm developers, with the most accurate algorithms producing many
fewer errors and undetectable false positive differentials. Since many
of the performance rates specified in the report would not be
acceptable for use in CBP operations, we do not use them.
---------------------------------------------------------------------------
\17\ Face Recognition Vendor Test (FRVT), Part 3: Demographic
Effects, National Institute of Standards and Technology, U.S.
Department of Commerce (December 2019), p.8.
---------------------------------------------------------------------------
CBP is committed to implementing the biometric entry exit mandate
in a way that provides a secure and streamlined travel experience for
all travelers, and CBP will continue to partner with NIST and use NIST
research to ensure the continued optimal performance of the CBP face
comparison service. In the upcoming weeks, CBP will directly provide
NIST with data for NIST to perform an independent and comprehensive
scientific analysis of CBP's operational face-matching performance,
including impacts due to traveler demographics and image quality. NIST
will provide objective recommendations regarding matching algorithms,
optimal thresholds, and gallery creation.
data security
There are 4 primary safeguards to secure passenger data, including
secure encryption during data storage and transfer, irreversible
biometric templates, brief retention periods, and secure storage.
Privacy is implemented by design, ensuring data protection through the
architecture and implementation of the biometric technology. CBP
prohibits its approved partners such as airlines, airport authorities,
or cruise lines from retaining the photos they collect as part of the
entry/exit program for their own business purposes. The partners must
immediately purge the images following transmittal to CBP, and the
partner must allow CBP to audit compliance with this requirement. As
discussed in its comprehensive November 2018 Privacy Impact Assessment
concerning its facial recognition technology, CBP has developed
business requirements, or system-wide standards, to document this
commitment.\18\ Our private-sector partners must agree as a condition
of participation in the pilots.
---------------------------------------------------------------------------
\18\ DHS/CBP (November 2018), DHS/CBP/PIA-056 Traveler Verification
Service. (945.31 KB).
---------------------------------------------------------------------------
privacy, transparency, civil rights, and future assessments
CBP is committed to ensuring that our use of technology sustains
and does not erode privacy protections. We take privacy very seriously
and are dedicated to protecting the privacy of all travelers. CBP
complies with the requirements of the Privacy Act of 1974 and all DHS
and Government-wide policies.\19\ In accordance with DHS policy, CBP
uses the Fair Information Practice Principles, or FIPPs, to assess the
privacy risks and ensure appropriate measures are taken to mitigate
risks from data collection through the use of biometrics. Our
partnering stakeholders are also held to the same standards.
---------------------------------------------------------------------------
\19\ Records maintained on individuals, 5 U.S.C. 552(a), P.L. 93-
579, 88 Stat. 1896.
---------------------------------------------------------------------------
CBP strives to be transparent and provide notice to individuals
regarding the collection, use, dissemination, and maintenance of
personally identifiable information (PII). When airlines or airports
partner with CBP on biometric air exit, the public is informed that the
partner is collecting the biometric data in coordination with CBP. We
notify travelers at these ports using verbal announcements, signs, and/
or message boards that CBP takes photos for identity verification
purposes, and they are informed of their ability to opt out. Foreign
nationals may opt out of providing biometric data to a third party, and
any U.S. citizen or foreign national may do so at the time of boarding
by notifying the airline-boarding agent that they would like to opt
out. The airline would conduct manual identity verification using their
travel document, and may notify CBP to collect biometrics, if
applicable.
If requested, CBP Officers provide a tear sheet with Frequently
Asked Questions, opt-out procedures, and additional information,
including the legal authority and purpose for inspection, the routine
uses, and the consequences for failing to provide information. CBP also
posts signs informing individuals of possible searches, and the purpose
for those searches, upon arrival or departure from the United States.
CBP provides general notification of its biometric exit efforts and
various pilot programs through Privacy Impact Assessments (PIAs) and
Systems of Records Notices (SORNs) and through information such as
Frequently Asked Questions, which are readily available at
www.cbp.gov.\20\
---------------------------------------------------------------------------
\20\ SORNs associated with CBP's Traveler Verification Service are:
DHS/CBP-007 Border Crossing Information, DHS/CBP-021 Arrival and
Departure Information System, DHS/CBP-006 Automated Targeting System,
DHS/CBP-011 U.S. Customs and Border Protection TECS. https://
www.dhs.gov/system-records-notices-sorns.
---------------------------------------------------------------------------
CBP published a comprehensive PIA concerning its facial recognition
technology, known as the Traveler Verification Service, in November
2018. An appendix to that document, published on January 8, 2020,
explains aspects of CBP's biometric use as well as policies and
procedures for the collection, storage, analysis, use, dissemination,
retention, and/or deletion of data.\21\ The PIA and the public notices
specifically highlight that facial images for arriving and departing
foreign nationals (and those dual national U.S. citizens traveling on
foreign documentation) are retained by CBP for up to 2 weeks, not only
to confirm travelers' identities but also to assure continued accuracy
of the algorithms and ensure there are no signs of any differential
performance. As always, facial images of arriving and departing foreign
nationals are forwarded to the IDENT system for future law enforcement
purposes, consistent with CBP's authority. As U.S. citizens are not
within the scope for biometric exit, photos of U.S. citizens used for
biometric matching purposes are held in secure CBP systems for no more
than 12 hours after identity verification in case of an extended system
outage or for disaster recovery.\22\ CBP reduced the retention period
for U.S. citizen photos to no more than 12 hours as a direct result of
briefings and consultations with Chairman Thompson.
---------------------------------------------------------------------------
\21\ DHS/CBP (November 2018), DHS/CBP/PIA-056 Traveler Verification
Service. (945.31 KB).
\22\ Controls of aliens departing from the United States;
Electronic visa update system, 8 C.F.R. 215; Inspection of persons
applying for admission, 8 C.F.R. 235.
---------------------------------------------------------------------------
CBP is committed to transparency in this process as well as to
improving its public messaging to help the public better understand the
technology. We welcome the committee's input. CBP collaborates
regularly with the DHS Privacy Office to ensure compliance with privacy
laws and policies. The DHS Privacy Office commissioned the DHS Data
Privacy and Integrity Advisory Committee (DPIAC) to advise the
Department on best practices for the use of facial comparison
technology. The DPIAC published its report on February 26, 2019.\23\
CBP has implemented or is actively working to implement all of the
DPIAC recommendations. CBP continues outreach efforts with privacy
advocacy groups regarding the biometric entry-exit program, most
recently meeting with them in December 2019. CBP also hosted the
Privacy and Civil Liberties Oversight Board (PCLOB) for a tour of
biometric processes at Atlanta/Hartsfield International Airport on
January 15, 2020.\24\
---------------------------------------------------------------------------
\23\ Report 2019-01 of the DHS Data Privacy and Integrity Advisory
Committee (DPIAC): Privacy Recommendations in Connection with the Use
of Facial Recognition Technology, Privacy Recommendations in Connection
with the Use of Facial Recognition Technology.pdf.
\24\ The Privacy Civil Rights Oversight Board is an independent,
bipartisan agency within the Executive branch established by the
Implementing Recommendations of the 9/11 Commission Act, P.L. 110-53,
https://www.pclob.gov/. Nextgov, Inside the CBP-Build `Backbone' of
Atlanta's Biometric Terminal, (January 21, 2020) inside-cbp-built-
backbone-atlantas-biometric-terminal.
---------------------------------------------------------------------------
cbp's progress toward implementing a comprehensive biometric entry-exit
system
Biometric Entry-Exit in the Air Environment
Facial comparison technology is enhancing the arrivals process,
enabling more efficient and more secure clearance processes that
benefit airports, airlines, and travelers with shorter connection times
and standardized arrival procedures. It is an additional tool to reduce
imposter threat while increasing the integrity of the immigration
system. Since initiating the use of facial comparison technology in the
air environment on a trial basis, CBP has identified 7 imposters,
including 2 with genuine U.S. travel documents (passport or passport
card), using another person's valid travel documents to seek entry into
the United States.\25\
---------------------------------------------------------------------------
\25\ Updated January 7, 2020.
---------------------------------------------------------------------------
CBP is working toward full implementation of biometric exit in the
air to account for over 97 percent of departing commercial air
travelers from the United States. Stakeholder partnerships are critical
for implementing a biometric entry-exit system, and airports, airlines,
and CBP are collaborating to develop a process that meets our biometric
entry-exit mandate and airlines' business needs. These partnerships
help ensure that biometric entry-exit does not have a detrimental
impact on the air travel industry, and that the technology is useful
and affordable. Stakeholders have attested that using biometrics could
lead to faster boarding times, enhanced customer service, better use of
our CBP staffing, and faster flight clearance times on arrival.
Engagement with additional stakeholders on how they can be incorporated
into the comprehensive entry-exit system continues, and CBP is ready to
partner with any appropriate airline or airport that wishes to use
biometrics to expedite the travel process for its customers.
Biometric Entry-Exit in the Land Environment
In the land environment, there are often geographical impediments
to expanding exit lanes to accommodate adding lanes or CBP-staffed
booths. The biometric exit land strategy focuses on implementing an
interim exit capability while simultaneously investigating what is
needed to implement a comprehensive system over the long term.
Biometrically verifying travelers who depart at the land border will
close a gap in the information necessary to complete a nonimmigrant
traveler's record in CBP's Arrival and Departure Information System,
and will allow us an additional mechanism to better determine when
travelers who depart the United States via land have overstayed their
admission period. Given DHS's desire to implement the use of biometrics
without negatively affecting cross-border commerce, CBP plans to take a
phased approached to land implementation.
Facial comparison technology, similar to what is used in the air
environment has been deployed at entry operations at the Nogales and
San Luis POEs in Arizona and at the Laredo and El Paso POEs in Texas.
CBP plans to expand to additional locations along the Southern Border
in 2020. By using the facial comparison technology in the land
environment, CBP has identified 247 imposters, including 45 with
criminal records and 18 under the age of 18, attempting to enter the
United States. Additionally, CBP tested ``at speed'' facial biometric
capture camera technology on vehicle travelers.\26\ From August 2018 to
February 28, 2019, CBP conducted a technical demonstration on people
inside vehicles moving less than 20 miles per hour entering and
departing Anzalduas, Texas.
---------------------------------------------------------------------------
\26\ DHS/CBP (November 2018), DHS/CBP/PIA-056 Traveler Verification
Service (945.31 KB).
---------------------------------------------------------------------------
Biometric Entry-Exit in the Sea Environment
Similar to efforts in the air environment, CBP is partnering with
the cruise line industry to use facial biometric processing supported
by CBP's biometric comparison service in the debarkation points at
seaports.\27\ Automating identity verification allows us to shift
officer focus to core law enforcement functions and reallocate
resources from primary inspections to roving enforcement activities.
Currently, there are 7 sea entry sites and 5 major cruise lines that
are operating facial comparison cameras to confirm arriving passenger
identity on closed-loop cruises, which begin and end in the same city.
Cruise lines report passenger satisfaction feedback that indicate the
debarkation process is significantly better than feedback from vessels
not using the technology during debarkation. CBP continues engagement
with cruise lines and port authorities to expand the technology to
other businesses and locations.
---------------------------------------------------------------------------
\27\ Ibid.
---------------------------------------------------------------------------
conclusion
DHS, in collaboration with the travel industry, is assertively
moving forward in developing a comprehensive biometric exit system in
the land, air, and sea environments that replace manual identity checks
with facial comparison technology. Travelers are well aware that their
picture is being taken for facial comparison purposes, and they have
access to both basic and detailed information regarding CBP's
collection of biometric information. Not only is CBP Congressionally-
mandated to implement a biometric entry-exit system, such a system will
enhance CBP's ability to accomplish its mission: To safeguard America's
borders thereby protecting the public from dangerous people and
materials while enhancing the Nation's global economic competitiveness
by enabling legitimate trade and travel. CBP's collaborative biometric
efforts address the recommendations of The 9/11 Commission Report,
specifically, that security and protection should be shared among the
various travel checkpoints (ticket counters, gates, and exit controls):
``By taking advantage of them all, we need not depend on any one point
in the system to do the whole job.''\28\
---------------------------------------------------------------------------
\28\ The 9/11 Commission, The 9/11 Commission Report, pp. 385-386,
http://govinfo.library.unt.edu/911/report/911Report.pdf. (7.22MB).
Chairman Thompson. Thank you for your testimony.
I now recognize Mr. Mina to summarize his statement for 5
minutes.
STATEMENT OF PETER E. MINA, DEPUTY OFFICER FOR PROGRAMS AND
COMPLIANCE, OFFICE OF CIVIL RIGHTS AND CIVIL LIBERTIES, U.S.
DEPARTMENT OF HOMELAND SECURITY
Mr. Mina. Good morning. Chairman Thompson, Ranking Member
Rogers, and distinguished Members of the committee, thank you
for the opportunity to appear before you today to discuss the
Department of Homeland Security's use of facial recognition
technology.
DHS's commitment to nondiscrimination in law enforcement
and screening activities remains an important cornerstone of
our daily work to secure the homeland.
I would like to make 3 overarching points in my testimony
today.
First, the Office of Civil Rights and Civil Liberties has
been and continues to be engaged with the DHS operational
components to ensure use of facial recognition technology is
consistent with civil right and civil liberties, law, and
policy.
Second, operators, researchers, and civil rights policy
makers must work together to prevent algorithms from leading to
impermissible biases in the use of facial recognition
technology.
Third, facial recognition technology can serve as an
important tool to increase the efficiency and effectiveness of
the Department's public protection mission, as well as the
facilitation of lawful travel.
But it is vital that these programs utilize technology in a
way that safeguards our Constitutional rights and values.
Now, to achieve these 3 points, CRCL, No. 1 influences DHS
policies and programs throughout their life cycle.
No. 2, engages with Department offices and components in
the development of new policies and programs to ensure that
protection of civil rights and civil liberties is fully
integrated into their foundation.
No. 3, monitors operational execution and engages with
stakeholders in order to provide feedback regarding the impacts
and consequences of policies and programs.
Fourth and finally, we investigate complaints and make
recommendations to DHS components, such as complaints including
allegations of racial profiling or other impermissible bias.
CRCL recognizes the potential risks of impermissible bias
in facial recognition algorithms, as previously raised by this
committee, and supports rigorous testing and evaluation of
algorithms used in facial recognition systems to identify and
mitigate impermissible bias.
CRCL will continue to support the collaborative
relationship between the National Institute of Standards and
Technology, the DHS Science and Technology Directorate, the DHS
Office of Biometric and Identity Management, and DHS
components, including U.S. Customs and Border Protection, to
that end.
In carrying out its mission, CRCL advised DHS components
and Department offices by participating in enterprise-level
groups working on biometric and facial recognition issues.
Further, CRCL directly engages with DHS components. For
example, CRCL has regularly engaged CBP on the implementation
of facial recognition technology and its biometric entry and
exit program.
In particular, CRCL advised on policy and implementation of
appropriate accommodations for individuals wearing religious
headwear, for individuals with a sincere religious objection to
being photographed, and for individuals who may have a
significant injury or disability for whom taking photographs
may present challenges or not be possible.
As DHS's facial recognition program has matured and
evolved, CRCL will be collaborating directly with CBP, S&T, and
OBIM to address potential civil rights and civil liberties
impacts.
Further, CRCL will engage communities with CBP and DHS S&T
to both inform the public regarding CBP's facial recognition
programs and address potential concerns.
Finally, we will continue to evaluate any potential alleged
violations of civil rights or civil liberties in order to
further inform our policy advice and strengthen DHS's facial
recognition programs.
CRCL understands that successful and appropriate facial
recognition technology requires on-going oversight and quality
assurance, initial validation and regular revalidation, and a
close relationship between the users and oversight offices.
In this way it can be developed to work properly and
without impermissible bias when it achieves initial operating
capability and then continually throughout its entire project
life cycle.
At the same time, we will need to work with the operational
components to ensure that policies and practices evolve so that
the human part of the equation, the users, are also focused on
responsible deployment of this technology, working in a manner
that prevents impermissible bias in DHS activities.
Again, I thank you for the opportunity to appear before you
today, and I look forward to answering your questions.
[The prepared statement of Mr. Mina follows:]
Prepared Statement of Peter E. Mina
February 6, 2020
Chairman Thompson, Ranking Member Rogers, and distinguished Members
of the committee, thank you for the opportunity to appear before you to
discuss the Department of Homeland Security's (DHS) use of facial
recognition technology. DHS's commitment to nondiscrimination in law
enforcement and screening activities remains an important cornerstone
of our daily work to secure the homeland.
I would like to make three overarching points in my testimony
today: (1) The Office for Civil Rights and Civil Liberties (CRCL) has
been and continues to be engaged with the DHS operational components to
ensure use of facial recognition technology is consistent with civil
rights and civil liberties law and policy; (2) operators, researchers,
and civil rights policy makers must work together to prevent algorithms
from leading to racial, gender, or other impermissible biases in the
use of facial recognition technology; and (3) facial recognition
technology can serve as an important tool to increase the efficiency
and effectiveness of the Department's public protection mission, as
well as the facilitation of lawful travel, but it is vital that these
programs utilize this technology in a way that safeguards our
Constitutional rights and values. To that end, we welcome the
opportunity to work with DHS policy makers and operators, Congress,
academic, and other non-Governmental entities on these important
issues.
introduction
CRCL supports the DHS mission to secure the Nation while preserving
individual liberty, fairness, and equality under the law. Established
by the Homeland Security Act of 2002, CRCL's mission integrates civil
rights and civil liberties into all DHS activities by:
Promoting respect for civil rights and civil liberties in
policy development and implementation by advising Department
leadership and personnel, and State and local partners;
Communicating with individuals and communities whose civil
rights and civil liberties may be affected by Department
activities, informing them about policies and avenues of
remedy, and promoting appropriate attention within the
Department to their experiences and concerns;
Investigating and resolving civil rights and civil liberties
complaints filed by the public regarding Department policies or
activities, or actions taken by Department personnel; and
Leading the Department's equal employment opportunity
programs and promoting workforce diversity and merit system
principles.
CRCL is a DHS headquarters office, and the CRCL officer reports
directly to the Secretary of Homeland Security. CRCL works
collaboratively with, but independently of, the DHS operational
components, including U.S. Customs and Border Protection (CBP). CRCL's
work is not, with limited but important exceptions,\1\ remedial in
nature.
---------------------------------------------------------------------------
\1\ CRCL has remedial authority under Section 504 of the
Rehabilitation Act of 1973, as amended, which states, ``No otherwise
qualified individual with a disability in the United States . . .
shall, solely by reason of her or his disability, be excluded from the
participation in, be denied the benefits of, or be subjected to
discrimination under any program or activity receiving Federal
financial assistance or under any program or activity conducted by any
Executive agency. . . .'' 29 U.S.C. 794.
---------------------------------------------------------------------------
Pursuant to statutory authorities under 6 U.S.C. 345 and 42
U.S.C. 2000ee-1, CRCL is responsible for assisting the Department in
developing, implementing, and periodically reviewing policies and
procedures to ensure the protection of civil rights and civil
liberties, including in CBP and other component screening and vetting
programs.
In carrying out its statutory mission, CRCL influences DHS policies
and programs throughout their life cycle. CRCL seeks to engage with
Department offices and components in the development of new policies
and programs to ensure that protection of civil rights and civil
liberties are fully integrated into their foundations. As
implementation begins, CRCL monitors operational execution and engages
with stakeholders in order to provide feedback to Department and
component leadership regarding the impacts or consequences of policies
and programs. Finally, CRCL investigates complaints and makes
recommendations to DHS components, often related to the creation or
modification of policies, or changes to implementation, training,
supervision, or oversight. Such complaints include allegations of
racial profiling or other impermissible bias. It is important to note
that the DHS Office of Inspector General has the right of first refusal
to investigate allegations submitted to CRCL.
dhs's use of facial recognition technology and crcl's role in oversight
DHS currently uses facial recognition technology to support CBP's
Biometric Entry-Exit Program and is researching and testing this
technology to see if it can be deployed in other mission areas, such as
identity verification in Transportation Security Administration
passenger screening. A key goal of the Department's use of facial
recognition technology is identifying and eliminating, to the extent it
exists, any impermissible bias based on race and gender. In addition to
the strong civil rights and civil liberties interest in ensuring
equality of treatment, the DHS operational components have a compelling
interest in ensuring the accuracy of this or any tool that assists in
performing the mission. Improved accuracy and efficiency in the
Department's data systems results in better performance of all the DHS
missions they support.
DHS partnered with the National Institute of Standards and
Technology (NIST) on the assessment of facial recognition technologies
to improve data quality and integrity, and ultimately the accuracy of
the technology, as a means of eliminating such impermissible bias.
Currently, the DHS Office of Biometric Identity Management
(OBIM) is partnering with NIST to develop a face image quality
standard that will improve the accuracy and reliability of
facial recognition as it is employed at DHS.
CBP is partnering with NIST to analyze performance impacts
due to image quality and traveler demographics and providing
recommendations regarding match algorithms, optimal thresholds
for false positives, and the selection of photographs used for
comparison.
DHS knows that accuracy and reliability, and the resulting
operational value of facial recognition technology, varies depending on
how the technology is employed. Variables include the nature of the
mission supported, variations in the type and quality of the
photographs, environmental factors such as lighting, the manner in
which the match is made, and the type of computer processing, including
the nature of the algorithms, used to make a match.
Human factors also matter. Users need to be aware of how the
technology works, its strengths and weaknesses, and how they can ensure
the technology functions in a way that complies with all applicable
laws and DHS policy. In addition to being operational considerations,
these factors also directly affect the civil rights and civil liberties
of those individuals who encounter this DHS technology. In short, the
legal and civil rights and civil liberties policy implications of
facial recognition technology depend on how the technology is
implemented.
CRCL recognizes the potential risks of impermissible bias in facial
recognition algorithms, as previously raised by this committee. CRCL
supports rigorous testing and evaluation of algorithms used in facial
recognition systems to identify and mitigate impermissible bias. CRCL
will continue to support the collaborative relationship between NIST,
the DHS Science & Technology Directorate, OBIM, and DHS components to
that end.
crcl uses partnerships and data to look beyond the algorithm
As discussed above, CRCL seeks to ensure civil rights and civil
liberties protections are incorporated into Department and component
programs--including the policies and practices that guide DHS use of
facial recognition technology. Our contribution to DHS working groups
is one way we fulfill our mission and identify areas that may require
further engagement.
CRCL participates in DHS enterprise-level groups working on
biometric and facial recognition issues, including:
The DHS Executive Steering Committee for Biometric
Capabilities, which provides coordination and guidance to all
DHS and component-level programs that are developing or
providing biometric capabilities in support of DHS mission
objectives. The Steering Committee serves as a forum for cross-
component collaboration and the sharing of biometric
challenges, needs, concepts, best practices, plans and efforts;
and
The Joint Requirements Council's Counter Terrorism and
Homeland Threats Portfolio Team, which is made up of component
subject-matter experts from the key functional areas within the
Department that validate and prioritize requirements and
capability gaps, to include those relating to biometrics and
screening and vetting functions.
Another way in which we carry out our role in providing proactive
advice is through direct engagement with DHS components. For example,
CRCL has regularly engaged CBP on the implementation of facial
recognition technology in its Biometric Entry-Exit Program. We have
viewed live demonstrations of the technology at Dulles International
Airport and Hartsfield-Jackson Airport in Atlanta. In addition, we
reviewed and commented on internal procedures, as well as proposed
regulations. CRCL advised on policy and implementation of appropriate
accommodations for individuals wearing religious headwear (e.g.,
individuals whose headwear may need to be adjusted to take a
photograph), for individuals with a sincere religious objection to
being photographed, and for individuals who may have a significant
injury or disability and for whom taking photographs may present
challenges or not be possible. CRCL and the DHS Privacy Office also
work cooperatively with the components to address and mitigate issues
such as photograph retention and data sharing.
We fully anticipate continuing to provide advice and guidance on
DHS's facial recognition programs as they mature and evolve, whether it
is through one of the Department's enterprise-level groups or directly
with the operational components.
Supporting our advisory role on new or proposed policies or
programs, I would also like to highlight the distinctive way CRCL uses
the information and allegations we receive as part of our compliance
process. In addition to the opening of formal investigations into
allegations of civil rights or civil liberties violations, when CRCL
does not open an investigation on an allegation, we use the information
received to track issues and identify potential patterns of alleged
civil rights or civil liberties violations that may require further
review. For CBP vetting operations, this data is used to guide CRCL in
identifying which policies or programs warrant further investigation to
more closely examine potentially serious or systemic issues.
Additionally, CRCL shares data with components annually to provide
visibility into the civil rights matters CRCL has received, and
publishes data on complaints in the Annual and Semi-Annual Reports to
Congress.
crcl's continuing efforts to safeguard civil rights and civil liberties
in dhs's use of emerging technologies
CRCL recognizes that facial recognition technology and the
computing that enable it are emerging technologies. They require
intensive support from all entities involved--operators, NIST and other
researchers, and oversight offices such as CRCL--to ensure that they
are compliant with applicable law and policy, including civil rights
and civil liberties protections, in all phases of development and
deployment. We understand that successful and appropriate facial
recognition technology requires on-going oversight and quality
assurance, initial validation and regular re-validation, and a close
relationship between the users and oversight offices. In this way, it
can be developed to work properly and without impermissible bias when
it achieves initial operating capability, and then continually through
its entire project life cycle. At the same time, we will need to work
with the operational components to ensure that policies and practices
evolve, to ensure that the human part of the equation--the users--are
also focused on the responsible deployment of this technology, working
in a manner that consistently prevents impermissible bias in DHS
activities.
As these and future projects develop, CRCL will remain engaged with
advocates, technologists, experts, and Congress to ensure that civil
rights and civil liberties protections are effective and sufficient.
Again, I thank you for the opportunity to appear before you today,
and I look forward to answering your questions.
Chairman Thompson. Thank you also for your testimony.
I now recognize Dr. Romine to summarize his statement for 5
minutes.
STATEMENT OF CHARLES H. ROMINE, Ph.D., DIRECTOR OF THE
INFORMATION TECHNOLOGY LABORATORY, NATIONAL INSTITUTE OF
STANDARDS AND TECHNOLOGY
Mr. Romine. Thank you, Chairman Thompson, Ranking Member
Rogers, and Members of the committee.
I am Chuck Romine, the director of the Information
Technology Laboratory of the National Institute of Standards
and Technology, also known as NIST.
Thank you for the opportunity to appear before you today to
discuss NIST's role in standards and testing for facial
recognition technology.
In the areas of biometrics, NIST has been working with
public and private sectors since the 1960's. Biometric
technologies provide a means to establish or verify the
identity of humans based upon one or more physical or
behavioral characteristics.
Face recognition technology compares an individual's facial
features to available images for verification or identification
purposes. NIST's work improves the accuracy, quality,
usability, interoperability, and consistency of identity
management systems and ensures that U.S. interests are
represented in the international arena.
NIST's research has provided state-of-the-art technology
benchmarks and guidance to industry and to U.S. Government
agencies that depend upon biometrics recognition technologies.
NIST's face recognition vendor testing program, or FRVT,
provides technical guidance and scientific support for analysis
and recommendations for utilization of face recognition
technologies to various U.S. Government and law enforcement
agencies, including the FBI, DHS, CBP, and IARPA.
The NIST FRVT Interagency Report 8280 released in December
2019 quantified the accuracy of face recognition algorithms for
demographic groups defined by sex, age, and race or country of
birth for both one-to-one and one-to-many identification search
algorithms. It found empirical evidence for the existence of
demographic differentials in facial recognition algorithms that
NIST evaluated.
The report distinguishes between false positive and false
negative errors and notes that the impacts of errors are
application-dependent.
NIST conducted tests to quantify demographic differences
for 189 face recognition algorithms from 99 developers using 4
collections of photographs with 18.27 million images of 8.49
million people.
These images came from operational databases provided by
the State Department, the Department of Homeland Security, and
the FBI.
I will first address one-to-one verification applications.
There, false positive differentials are much larger than those
related to false negative and exist across many of the
algorithms tested. False positives might present a security
concern to the system owner as they may allow access to
imposters.
Other findings are that false positives are higher in women
than in men and are higher in the elderly and the young
compared to middle-aged adults.
Regarding race, we measured higher false positive rates in
Asian and African American faces relative to those of
Caucasians. There are also higher false positive rates in
Native Americans, American Indian, Alaskan Indian, and Pacific
Islanders.
These effects apply to most algorithms, including those
developed in Europe and the United States. However, a notable
exception was for some algorithms developed in Asian countries.
There was no such dramatic difference in false positives in
one-to-one matching between Asian and Caucasian faces for the
algorithms developed in Asia.
While the NIST study did not explore the relationship
between cause and effect, one possible connection and an area
for research is the relationship between algorithm's
performance and the data used to train the algorithm itself.
I will now comment on one-to-many search algorithms. Again,
the impact of errors is application-dependent. False positives
in one-to-many search are particularly important because the
consequences could include false accusations.
For most algorithms, the NIST study measured higher false
positive rates in women, African Americans, and particularly in
African American women. However, the study found that some one-
to-many algorithms gave similar false positive rates across
these specific demographics. Some of the most accurate
algorithms fell into this group.
This last point underscores one overall message of the
report: Different algorithms perform differently.
Indeed, all of our FRVT reports note wide variations in
recognition accuracy across algorithms, and an important result
from the demographic study is that demographic effects are
smaller with more accurate algorithms.
NIST is proud of the positive impact it has had in the last
60 years on the evolution of biometrics capabilities. With
NIST's extensive experience and broad expertise both in its
laboratories and in successful collaborations with the private
sector and other Government agencies, NIST is actively pursuing
the standards and measurement research necessary to deploy
interoperable, secure, reliable, and usable identity management
systems.
Thank you for the opportunity to testify on NIST's
activities in facial recognition and identity management, and I
would be happy to answer any questions you may have.
[The prepared statement of Dr. Romine follows:]
Prepared Statement of Charles H. Romine
February 6, 2020
introduction
Chairman Thompson, Ranking Member Rogers, and Members of the
committee, I am Chuck Romine, director of the Information Technology
Laboratory (ITL) at the Department of Commerce's National Institute of
Standards and Technology (NIST). ITL cultivates trust in information
technology and metrology through measurements, standards, and testing.
Thank you for the opportunity to appear before you today to discuss
NIST's role in standards and testing for facial recognition technology.
biometric and facial recognition technology
Home to 5 Nobel Prizes, with programs focused on National
priorities such as advanced manufacturing, the digital economy,
precision metrology, quantum science, and biosciences, NIST's mission
is to promote U.S. innovation and industrial competitiveness by
advancing measurement science, standards, and technology in ways that
enhance economic security and improve our quality of life.
In the area of biometrics, NIST has been working with public and
private sectors since the 1960's. Biometric technologies provide a
means to establish or verify the identity of humans based upon one or
more physical or behavioral characteristics. Examples of physical
characteristics include face, fingerprint, and iris images. An example
of behavioral characteristic is an individual's signature. Used with
other authentication technologies, such as passwords, biometric
technologies can provide higher degrees of security than other
technologies employed alone. For decades, biometric technologies were
used primarily in homeland security and law enforcement applications,
and they are still a key component of these applications. Over the past
several years, the marketplace for biometric solutions has widened
significantly and today includes public and private-sector applications
world-wide, including physical security, banking, and retail
applications. According to one industry estimate, the biometrics
technology market size will be worth $59.31 billion by 2025.\1\ There
has been a considerable rise in development and adoption of facial
recognition, detection, and analysis technologies in the past few
years.
---------------------------------------------------------------------------
\1\ https://www.grandviewresearch.com/industry-analysis/biometrics-
industry.
---------------------------------------------------------------------------
Face detection technology determines whether the image contains a
face. Face analysis technology aims to identify attributes such as
gender, age, or emotion from detected faces. Face recognition
technology compares an individual's facial features to available images
for verification or identification purposes. Verification or ``one-to-
one'' matching confirms a photo matches a different photo of the same
person in a database or the photo on a credential, and is commonly used
for authentication purposes, such as unlocking a smartphone or checking
a passport. Identification or ``one-to-many'' search determines whether
the person in the photo has any match in a database and can be used for
identification of a person.
Accuracy of face recognition algorithms is assessed by measuring
the two classes of error the software can make: False positives and
false negatives. A false positive means that the software wrongly
considered photos of 2 different individuals to show the same person,
while a false negative means the software failed to match 2 photos
that, in fact, do show the same person.
nist's role in biometric and facial recognition technology
NIST responds to Government and market requirements for biometric
standards, including facial recognition technologies, by collaborating
with other Federal agencies, law enforcement, industry, and academic
partners to:
research measurement, evaluation, and interoperability to
advance the use of biometric technologies including face,
fingerprint, iris, voice, and multi-modal techniques;
develop common models and metrics for identity management,
critical standards, and interoperability of electronic
identities;
support the timely development of scientifically valid, fit-
for-purpose standards; and
develop the required conformance testing architectures and
testing tools to test implementations of selected standards.
NIST's work improves the accuracy, quality, usability,
interoperability, and consistency of identity management systems and
ensures that United States interests are represented in the
international arena. NIST research has provided state-of-the-art
technology benchmarks and guidance to industry and to U.S. Government
agencies that depend upon biometrics recognition technologies.
Under the provisions of the National Technology Transfer and
Advancement Act of 1995 (Public Law 104-113) and OMB Circular A-119,
NIST is tasked with the role of encouraging and coordinating Federal
agency use of voluntary consensus standards in lieu of Government-
unique standards, and Federal agency participation in the development
of relevant standards, as well as promoting coordination between the
public and private sectors in the development of standards and in
conformity assessment activities. NIST works with other agencies to
coordinate standards issues and priorities with the private sector
through consensus standards developing organizations such as the
International Committee for Information Technology Standards (INCITS),
Joint Technical Committee 1 of the International Organization for
Standardization/International Electrotechnical Commission (ISO/IEC),
the Organization for the Advancement of Structured Information
Standards (OASIS), IEEE, the Internet Engineering Task Force (IETF),
and other standards organizations such as the International Civil
Aviation Organization (ICAO), and the International Telecommunication
Union's Standardization Sector (ITU-T). NIST leads National and
international consensus standards activities in biometrics, such as
facial recognition technology, but also in cryptography, electronic
credentialing, secure network protocols, software and systems
reliability, and security conformance testing--all essential to
accelerate the development and deployment of information and
communication systems that are interoperable, reliable, secure, and
usable.
Since 2010, NIST has organized the biennial International Biometric
Performance Testing Conference. This series of conferences accelerates
adoption and effectiveness of biometric technologies by providing a
forum to discuss and identify fundamental, relevant, and effective
performance metrics, and disseminating best practices for performance
design, calibration, evaluation, and monitoring.
facial recognition tests and evaluations
For more than a decade, NIST biometric evaluations have measured
the core algorithmic capability of biometric recognition technologies
and reported the accuracy, throughput, reliability, and sensitivity of
algorithms with respect to data characteristics, for example, noise or
compression, and to subject characteristics, for example, age or
gender. NIST biometric evaluations advance the technology by
identifying and reporting gaps and limitations of current biometric
recognition technologies. NIST evaluations advance measurement science
by providing a scientific basis for ``what to measure'' and ``how to
measure.'' NIST evaluations also facilitate development of consensus-
based standards by providing quantitative data for development of
scientifically sound, fit-for-purpose standards.
NIST conducted the Face Recognition Grand Challenge (2004-2006) and
Multiple Biometric Grand Challenge (2008-2010) programs to challenge
the facial recognition community to break new ground solving research
problems on the biometric frontier.
Since 2000, NIST's Face Recognition Vendor Testing Program (FRVT)
has assessed capabilities of facial recognition algorithms for one-to-
many identification and one-to-one verification. Participation in FRVT
is open to any organization world-wide. There is no charge for
participation, and being an on-going activity, participants may submit
their algorithms on a continuous basis. The algorithms are submitted to
NIST by corporate research and development laboratories and a few
universities. As prototypes, these algorithms are not necessarily
available as mature integrable products. For all algorithms that NIST
evaluates, NIST posts performance results on its FRVT website and
identifies the algorithm and the developing organization.
NIST and the FRVT program do not train face recognition algorithms.
NIST does not provide training data to the software under test, and the
software is prohibited from adapting to any data that is passed to the
algorithms during a test.\2\
---------------------------------------------------------------------------
\2\ The process of training a face recognition algorithm (or any
machine learning algorithm) involves providing a machine learning
algorithm with training data to learn from. The training data shall
contain the correct answer, which is known as ground-truth label, or a
target. The learning algorithm finds patterns in the training data that
map the input data attributes to the target and builds a machine-
learning model that captures these patterns. This model can then be
used to get predictions on new data for which the target is unknown.
---------------------------------------------------------------------------
NIST provides technical guidance and scientific support for
analysis and recommendations for utilization of facial recognition
technologies to various U.S. Government and law enforcement agencies,
including the Federal Bureau of Investigation (FBI), Office of
Biometric Identity Management (OBIM) at the Department of Homeland
Security (DHS), Department of Homeland Security Science and Technology
Directorate (DHS S&T), the Department of Homeland Security's U.S.
Customs and Border Protection agency (DHS CBP), and the Intelligence
Advanced Research Projects Activity (IARPA) at the office of the
Director of National Intelligence.
Historically and currently, NIST biometrics research has assisted
DHS. For example, NIST's research was used by DHS in its transition to
ten prints for the former US-VISIT program. NIST is currently
collaborating with DHS OBIM on face image quality standards.
Additionally, NIST is working with DHS CBP to analyze performance
impacts due to image quality and traveler demographics and provide
recommendations regarding match algorithms, optimal thresholds and
match gallery creation for its Traveler Verification Service (TVS).
nist face recognition vendor testing program
NIST's Face Recognition Vendor Testing Program (FRVT) was
established in 2000 to provide independent evaluations of both
prototype and commercially-available facial recognition algorithms.
These evaluations provide the U.S. Government with information to
assist in determining where and how facial recognition technology can
best be deployed. FRVT results also help identify future research
directions for the facial recognition community.
The 2013 FRVT tested facial recognition algorithms submitted by 16
organizations, and showed significant algorithm improvement since
NIST's 2010 FRVT test. NIST defined performance by recognition
accuracy--how many times the software correctly identified the photo--
and the time the algorithms took to match one photo against large photo
data sets.
The 2018 FRVT tested 127 facial recognition algorithms from the
research laboratories of 39 commercial developers and one university,
using 26 million mugshot images of 12 million individuals provided by
the FBI. The 2018 FRVT measured the accuracy and speed of one-to-many
facial recognition identification algorithms. The evaluation also
contrasted mugshot accuracy with that from lower quality images. The
findings, reported in NIST Interagency Report 8238,\3\ showed that
massive gains in accuracy have been achieved since the FRVT in 2013,
which far exceed improvements made in the prior period (2010-2013). The
accuracy gains observed in the 2018 FVRT study stem from the
integration, or complete replacement, of older facial recognition
techniques with those based on deep convolutional neural networks.
While the industry gains are broad, there remains a wide range of
capabilities, with some developers providing much more accurate
algorithms than others. Using FBI mugshots, the most accurate
algorithms fail only in about \1/4\ of 1 percent of searches, and these
failures are associated with images of injured persons and those with
long time lapse since the first photograph. The success of mugshot
searches stems from the new generation of facial recognition
algorithms, and from the adoption of portrait photography standards
first developed at NIST in the late 1990's.
---------------------------------------------------------------------------
\3\ https://nvlpubs.nist.gov/nistpubs/ir/2018/NIST.IR.8238.pdf.
---------------------------------------------------------------------------
The 2019 FRVT quantified the accuracy of face recognition
algorithms for demographic groups defined by sex, age, and race or
country of birth, for both one-to-one verification algorithms and one-
to-many identification search algorithms. NIST conducted tests to
quantify demographic differences for 189 face recognition algorithms
from 99 developers, using 4 collections of photographs with 18.27
million images of 8.49 million people. These images came from
operational databases provided by the State Department, the Department
of Homeland Security, and the FBI. Previous FRVT reports \4\ documented
the accuracy of these algorithms and showed a wide range in accuracy
across algorithms. The more accurate algorithms produce fewer errors
and can therefore be anticipated to have smaller demographic
differentials.
---------------------------------------------------------------------------
\4\ Part 1: https://www.nist.gov/system/files/documents/2019/11/20/
frvt_report_2019- _11_19_0.pdf and Part 2: https://www.nist.gov/system/
files/documents/2019/09/11/ni- stir_8271_20190911.pdf.
---------------------------------------------------------------------------
NIST Interagency Report 8280,\5\ released on December 19, 2019,
quantifies the effect of age, race, and sex on face recognition
performance. It found empirical evidence for the existence of
demographic differentials in face recognition algorithms that NIST
evaluated. The report distinguishes between false positive and false
negative errors, and notes that the impacts of errors are application
dependent.
---------------------------------------------------------------------------
\5\ https://nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8280.pdf.
---------------------------------------------------------------------------
I will first address one-to-one verification applications. There,
false positive differentials are much larger than for false negatives
and exist across many, but not all, algorithms tested. Across
demographics, false positives rates often vary by factors of 10 to
beyond 100 times. False negatives tend to be more algorithm-specific,
and often vary by factors below 3. False positives might present a
security concern to the system owner, as they may allow access to
impostors. False positives may also present privacy and civil rights
and civil liberties concerns such as when matches result in additional
questioning, surveillance, errors in benefit adjudication, or loss of
liberty. False positives are higher in women than in men and are higher
in the elderly and the young compared to middle-aged adults. Regarding
race, we measured higher false positive rates in Asian and African
American faces relative to those of Caucasians. There are also higher
false positive rates in Native American, American Indian, Alaskan
Indian, and Pacific Islanders. These effects apply to most algorithms,
including those developed in Europe and the United States. However, a
notable exception was for some algorithms developed in Asian countries.
There was no such dramatic difference in false positives in one-to-one
matching between Asian and Caucasian faces for algorithms developed in
Asia. While the NIST study did not explore the relationship between
cause and effect, one possible connection, and area for research, is
the relationship between an algorithm's performance and the data used
to train the algorithm itself.
I will now comment on one-to-many search algorithms. Again, the
impact of errors is application-dependent. False positives in one-to-
many search are particularly important because the consequences could
include false accusations. For most algorithms, the NIST study measured
higher false positives rates in women, African Americans, and
particularly in African American women. However, the study found that
some one-to-many algorithms gave similar false positive rates across
these specific demographics. Some of the most accurate algorithms fell
into this group. This last point underscores one overall message of the
report: Different algorithms perform differently. Indeed all of our
FRVT reports note wide variations in recognition accuracy across
algorithms, and an important result from the demographics study is that
demographic effects are smaller with more accurate algorithms.
A general takeaway from these studies is that, there is significant
variance between the performance facial recognition algorithms, that
is, some produce significantly fewer errors than others. Consequently,
users, policy makers, and the public should not think of facial
recognition as either always accurate or always error prone.
nist face in video evaluation program
The Face in Video Evaluation Program (FIVE) assessed the capability
of facial recognition algorithms to correctly identify or ignore
persons appearing in video sequences. The outcomes of FIVE are
documented in NIST Interagency report 8173,\6\ which enumerates
accuracy and speed of facial recognition algorithms applied to the
identification of persons appearing in video sequences drawn from 6
different video datasets. NIST completed this program in 2017.
---------------------------------------------------------------------------
\6\ https://www.nist.gov/publications/face-video-evaluation-five-
face-recognition-non-cooperative-subjects.
---------------------------------------------------------------------------
human factors: facial forensic examiners
NIST is researching how to measure the accuracy of forensic
examiners matching identity across different photographs. The study
measures face identification accuracy for an international group of
professional forensic facial examiners working under circumstances
approximating real-world casework. The findings, published in the
proceedings of the National Academy of Sciences,\7\ showed that
examiners and other human face ``specialists,'' including forensically-
trained facial reviewers and untrained super-recognizers, were more
accurate than the control groups on a challenging test of face
identification. It also presented data comparing state-of-the-art
facial recognition algorithms with the best human face identifiers. The
best machine performed in the range of the best-performing humans, who
were professional facial examiners. However, optimal face
identification was achieved only when humans and machines collaborated.
---------------------------------------------------------------------------
\7\ https://www.pnas.org/content/115/24/6171.
---------------------------------------------------------------------------
voluntary consensus standards
When properly conducted, standards development can increase
productivity and efficiency in Government and industry, expand
innovation and competition, broaden opportunities for international
trade, conserve resources, provide consumer benefit and choice, improve
the environment, and promote health and safety.
In the United States, most standards development organizations are
industry-led private-sector organizations. Many voluntary consensus
standards from those standard development organizations are appropriate
or adaptable for the Government's purposes. OMB Circular A-119 directs
the use of such standards by U.S. Government agencies, whenever
practicable and appropriate, to achieve the following goals:
eliminating the cost to the Federal Government of developing
its own standards and decreasing the cost of goods procured and
the burden of complying with agency regulation;
providing incentives and opportunities to establish
standards that serve National needs, encouraging long-term
growth for U.S. enterprises and promoting efficiency, economic
competition, and trade; and
furthering the reliance upon private-sector expertise to
supply the Federal Government with cost-efficient goods and
services.
examples of nist consensus standards development activities
ANSI/NIST-ITL.--The ANSI/NIST-ITL standard for biometric
information is used in 160 countries to ensure biometric data exchange
across jurisdictional line and between dissimilar systems. One of the
important effects of NIST work on this standard is that it allows
accurate and interoperable exchange of biometrics information by law
enforcement globally and enables them to identify criminals and
terrorists. NIST's own Information Technology Laboratory is an American
National Standards Institute (ANSI)-accredited standard development
organization. Under accreditation by ANSI, the private-sector U.S.
standards federation, NIST continues to develop consensus biometric
data interchange standards. Starting in 1986, NIST has developed and
approved a succession of data format standards for the interchange of
biometric data. The current version of this standard is ANSI/NIST-ITL
1: 2015, Data Format for the Interchange of Fingerprint, Facial & Other
Biometric Information.\8\ This standard continues to evolve to support
Government applications including law enforcement, homeland security,
as well as other identity management applications. Virtually all law
enforcement biometric collections world-wide use the ANSI/NIST-ITL
standard. NIST biometric technology evaluations in fingerprint, face,
and iris have provided the Government with timely analysis of market
capabilities to guide biometric technology procurements and
deployments.
---------------------------------------------------------------------------
\8\ https://www.nist.gov/publications/data-format-interchange-
fingerprint-facial-other-biometric-information-ansinist-itl-1-1.
---------------------------------------------------------------------------
iso/iec joint technical committee 1, subcommittee 37 (jtc1/sc37)--
biometrics
From the inception of the ISO Subcommittee on Biometrics in 2002,
NIST has led and provided technical expertise to develop international
biometric standards in this subcommittee. Standards developed by the
Subcommittee on Biometrics have received wide-spread international and
National market acceptance. Large international organizations, such as
the ICAO for Machine-Readable Travel Documents and the International
Labour Office (ILO) of the United Nations for the verification and
identification of seafarers, specify in their requirements the use of
some of the international biometric standards developed by this
subcommittee.
Since 2006, JTC1/SC37 has published a series of standards on
biometric performance testing and reporting, many of which are based on
NIST technical contributions. These documents provide guidance on the
principles and framework, testing methodologies, modality-specific
testing, interoperability performance testing, access control
scenarios, and testing of on-card comparison algorithms for biometric
performance testing and reporting. NIST contributes toward the
development of these documents and follows their guidance and metrics
in its evaluations, such as the FRVT.
conclusion
NIST is proud of the positive impact it has had in the last 60
years on the evolution of biometrics capabilities. With NIST's
extensive experience and broad expertise, both in its laboratories and
in successful collaborations with the private sector and other
Government agencies, NIST is actively pursuing the standards and
measurement research necessary to deploy interoperable, secure,
reliable, and usable identity management systems.
Thank you for the opportunity to testify on NIST's activities in
facial recognition and identity management. I would be happy to answer
any questions that you may have.
Chairman Thompson. Thank you very much.
I thank all of the witnesses for their testimony.
I remind each Member that he or she will have 5 minutes to
question the panel.
I will now recognize myself for questions.
Dr. Romine, we will start off with you. Part of your NIST
report was like next generation technology, as I understand,
that CBP will use or did you review existing technology?
Mr. Romine. We are not certain of that. We certainly intend
to continue our investigations. The existence of the specific
algorithms that we test, those algorithms are submitted to us
by the vendors. We have no independent way to correlate whether
those are the identical algorithms that are being used in the
field.
Chairman Thompson. So part of what you said is how the
technology is deployed depends on the application of the
technology. Explain that a little more to the committee.
Mr. Romine. Certainly. Our approach is that the significant
thing to be cognizant of is the risk associated with the
deployment, and the studies that we do help to inform policy
makers, such as Members of Congress, as well as operators of
these technologies, about how to quantify those risks at least
for the algorithms themselves.
The deployed systems have other characteristics associated
with them that we do not test. We test only the algorithms
currently.
The second point is that that risk that comes from the
error rates associated with the algorithms is part of a much
larger risk management that the operators have to undertake.
For example, access to critical infrastructures and access
control systems to critical infrastructures is different than
access to a phone that you might have. The risks are different
in those cases.
Chairman Thompson. Thank you.
Mr. Wagner, can you share with the committee the extent
that CBP goes to to protect the information collected in this
process?
Mr. Wagner. Sure. So the photographs that are taken by one
of our stakeholders' cameras, they are encrypted. They are
transmitted securely to the CBP cloud infrastructure where the
gallery is positioned.
Chairman Thompson. Right.
Mr. Wagner. The pictures are templatized, which means they
are turned into some type of mathematical structure that cannot
be reverse-engineered, and they are matched up with the
templatized photos that we have pre-staged in the gallery, and
then just a response goes back, yes or no, with a unique
identifier.
Chairman Thompson. Thank you.
So the comment that 2 to 3 percent of people who are
misidentified, what is CBP doing to try to get that to zero?
Mr. Wagner. Right. So it is not that they are
misidentified. It just means we did not match them to a picture
in the gallery that we did have of them. So we should have
matched them.
You are right. That should be at zero, and that is where we
look at the operational variables, the camera, the picture
quality, the human behaviors when the photo was taken, the
lighting, those different types, and then the age of the photo.
Then what we have seen in the NIST report, your gallery
size impacts your match rate. I think NIST tested galleries up
to 12 million size. We are comparing against a few thousand
here at most.
Then the numbers of photos that we have of the particular
individual can impact which one we match against and then some
of your match rates, and then the age of the photo. So if you
had your passport taken at age 20 and you are now 29 and your
face has changed in the dimensions, we are going to struggle to
match against that, which is then compounded by poor lighting
conditions or the person moving when the photo is taken or a
poorer-quality photo.
Chairman Thompson. Well, Mr. Mina, listening to what you
just heard, have you all dealt with any complaints from
citizens about this technology?
Mr. Mina. Mr. Chairman, we have received one complaint that
referenced this facial recognition technology. However, we have
not seen a trend, and that is when we would actually, in fact,
open an investigation in this matter.
We are working, as I mentioned, on the policy side of the
house advising CBP directly.
The other way in which we also hear from the community, as
you may know, is through our Community Engagement Roundtables
around the country, and we have heard concerns in those forums
about facial recognition technology, and those are concerns
that we are using to inform our advice.
Chairman Thompson. So can you provide the committee with
where you have held those forums around the country?
Mr. Mina. Yes, absolutely.
We do roundtables in about 18 cities, and not to say that
these concerns have been raised in every single location, but
certainly in some.
Then, again, we will continue to have those discussions
with CBP and with S&T in the future at future roundtables.
Chairman Thompson. Thank you.
Last, Mr. Wagner, I am not sure you have information on
this, but last month Iranian and Lebanese nationals and
individuals who travel to Iran and Lebanon, most of whom were
U.S. citizens or green card holders, were targeted, detained,
and subjected to prolonged questioning of up to 12 hours at the
Blaine area port of entry. I understand an internal CBP memo
indicates people were also questioned based on their religion,
which is completely unacceptable.
I understand CBP has admitted to enormous mistakes in this
incident. If you know, how did this situation happen?
What is CBP doing to ensure that it never happens again?
Mr. Wagner. So there was no National directive or guidance
that went out other than because of the things taking place in
Iran, the concerns about retaliation, we put our field managers
on alert to be more vigilant about current events that are
happening and work with your State, local, and Federal
counterparts and, you know, really just be vigilant.
There was some more prescriptive guidance that went out at
the local level in Blaine, Washington, which we are reviewing
right now because there are a lot of concerning things, I
think, that we saw in the interpretation of that guidance and
the management oversight as that weekend was unfolding and
people were being referred in for additional inspections and
questioning, and there are some concerning points about the
management engagement or lack thereof of what transpired.
So there is an internal investigation that CBP is
conducting. Civil Rights and Civil Liberties is conducting an
investigation, and when we get the results of that, we will
then proceed, you know, accordingly, depending on what those
results say.
Chairman Thompson. Mr. Mina, were you aware of that?
Mr. Mina. Yes, and as Mr. Wagner said, we do have an open
investigation in this matter.
Chairman Thompson. OK. Thank you.
Ranking Member, are you ready?
Mr. Rogers. I am ready. Thank you.
Chairman Thompson. I yield to the Ranking Member for an
opening statement.
Mr. Rogers. I am sorry for being late. We just got back
from the National Prayer Breakfast.
Thank you, Mr. Chairman.
After the tragic events of September 11, Congress
recognized that biometric systems are essential to our homeland
security. Following the recommendation of the 9/11 Commission,
Congress charged DHS with the creation of an automated,
biometric entry and exit system.
Customs and Border Protection and the Transportation and
Security Administration have already demonstrated the
capability of biometrics to improve security, facilitate
travel, and better enforced existing immigration law.
Government and the private sector have made enormous
strides in the accuracy, speed, and deployment of biometric
systems. Biometric technologies of all types have seen
improvements.
These advances in facial recognition algorithms, in
particular, are transformational. The National Institute of
Standards and Technology is the leader in testing and
evaluation for biometric technologies.
Dr. Romine and his team have done incredible work to help
Congress, DHS, and industry understands the capability of
currently available algorithms, but I am concerned that some of
my colleagues have already jumped to the misleading conclusion
that NIST reports on facial recognition.
Just hours after NIST released the 1,200 pages of technical
data, the Majority tweeted that this report shows facial
recognition is even more unreliable and racially biased than we
feared. If the Majority had taken the time to read the full
report before tweeting, they would have found that the real
headline, NIST determined that facial recognition algorithms
being adopted by DHS has no statistically detectable race or
gender bias.
In other words, NIST could find no statistical evidence
that facial recognition algorithms that DNS is adopting
contains racial bias.
I hope my colleagues will listen to Dr. Romine as he
explains how the NIST report proves that race or gender bias is
statistically undetectable in the most accurate algorithms.
The reality is that facial recognition technologies can
improve existing processes by reducing human error. These
technologies are tools that cannot and will not replace the
final judgment of CBP or TSA officers.
Concerns regarding privacy and civil rights are well-
intentioned, but these concerns can be fully addressed in how
biometric systems are implemented by DHS.
I look forward to hearing the steps that CRCL is taking to
coordinate with CBP and to protect privacy and civil rights of
Americans.
But as I have said before, halting all Government biometric
programs is not a solution. Doing so ignores the critical facts
that the technology that DHS uses is not racially biased. It
does not violate the civil rights of Americans. It is accurate.
Most importantly, it does protect the homeland.
I appreciate the Chairman calling the hearing today. It is
important for Congress to further educate itself on this issue.
I look forward to getting the facts, and I yield back, Mr.
Chairman.
[The statement of Ranking Member Rogers follows:]
Statement of Ranking Member Mike Rogers
February 6, 2020
After the tragic events of September 11, Congress recognized that
biometric systems are essential to our homeland security.
Following the recommendation of the 9/11 Commission, Congress
charged DHS with the creation of an automated biometric entry and exit
system.
Customs and Border Protection and the Transportation Security
Administration have already demonstrated the capability of biometrics
to improve security, facilitate travel, and better enforce existing
immigration laws.
Government and the private sector have made enormous strides in the
accuracy, speed, and deployment of biometrics systems.
Biometric technologies of all types have seen improvements.
The advances in facial recognition algorithms in particular are
transformational.
The National Institute of Standards and Technology is the leader in
testing and evaluation for biometric technologies.
Dr. Romine and his team have done incredible work to help Congress,
DHS, and industry understand the capability of currently-available
algorithms.
But I'm concerned that some of my colleagues have already jumped to
misleading conclusions regarding the NIST report on facial recognition.
Just hours after NIST released over 1,200 pages of technical data,
the Majority tweeted ``This report shows facial recognition is even
more unreliable and racially biased than we feared . . . [''.
If the Majority had taken the time to read the full report before
tweeting, they would have found the real headline: NIST determined that
the facial recognition algorithm being adopted by DHS had no
statistically detectable race or gender bias.
In other words, NIST could find NO statistical evidence that the
facial recognition algorithm DHS is adopting contains racial bias.
NIST found measurable and significant errors and bias in OTHER
facial recognition algorithms, but NOT in the algorithm used by DHS.
I hope that my colleagues will listen when Dr. Romine explains how
the NIST report proves that race or gender bias is statistically
undetectable in the most accurate algorithms.
The reality is that facial recognition technologies can improve
existing processes by reducing human error.
These technologies are tools that cannot and will not replace the
final judgment of CBP or TSA officers.
Concerns regarding privacy and civil rights are well-intentioned.
But these concerns can be fully addressed in how biometric systems
are implemented by DHS.
I look forward to hearing the steps CRCL is taking to coordinate
with CBP and protect the privacy and civil rights of Americans.
But as I have said before, halting all Government biometric
programs is not the solution.
Doing so ignores these critical facts: The technology DHS uses is
NOT racially biased; It does NOT violate the civil rights of Americans;
It IS accurate; and most importantly, it DOES protect the homeland.
I appreciate the Chairman calling this hearing today. It's
important for Congress to further educate itself on this issue. I look
forward to getting the facts on the record.
Chairman Thompson. Thank you very much.
I wish you had heard the testimony because there was some
testimony we heard to the contrary.
Mr. Rogers. I look forward to probing them on that.
Chairman Thompson. All right. Well, I recognize the
gentleman for his questions.
Mr. Rogers. My statement is wrong, to get to the Chairman's
point. Anybody can jump at it.
Mr. Wagner. I would never tell Congress they are wrong.
Mr. Rogers. You are one of the few people who will not do
that.
[Laughter.]
Mr. Rogers. Literally, I mean, my understanding is there is
no statistical evidence that there is racial bias. Is that an
inaccurate statement?
Mr. Romine. Thank you for the question.
In the highest-performing algorithms for one-to-many
matches, the highest-performing algorithms we saw undetectable
bias. The demographic differentials that we were measuring we
say are undetectable in the report.
Mr. Rogers. So what do you mean by undetectable?
Mr. Romine. What I mean by that is that in the testing that
we undertook, there was no way to determine--let me back up and
say the idea of having absolutely zero false positives is a big
challenge.
Mr. Rogers. Did you test the NEC-3 algorithm being used by
DHS?
Mr. Romine. We tested algorithms from NEC. We have no
independent way to verify that that is the specific algorithm
that is being used by CBP. That would be something that CBP and
NEC would have to attest to.
From our perspective, the vendor provides us algorithms.
They are black boxes that we test just the performance of the
algorithm that is submitted to us by the vendor.
Mr. Rogers. Mr. Wagner, is CBP currently working to
implement NEC-3 algorithms?
Mr. Wagner. Any what? I am sorry.
Mr. Rogers. NEC-3 algorithms.
Mr. Wagner. Yes, we are using an earlier version of NEC
right now, and I believe we are testing NEC-3, which is the
version that was tested, and the plan is to use it next month
in March to switch or to upgrade basically to that one.
Mr. Rogers. OK. Dr. Romine, who can participate in the
facial recognition vendor test? Is it accurate to say that some
algorithms are far less accurate and sophisticated than others?
Mr. Romine. Yes, sir, that is correct. Anyone around the
globe can participate. We have participants from industries,
biometrics industries around the country, but also from
universities and some experimental systems as well.
Mr. Rogers. Great. That is all I have, Mr. Chairman. Thank
you.
Chairman Thompson. Thank you very much.
Mr. Wagner, let's get clear. The C-3, you do not have it
operational anywhere in the country, right? You are testing it.
That technology goes into being, you said, next month?
Mr. Wagner. The NEC-3 algorithm we are planning to
implement next month. The earlier version of it is operational
now.
Chairman Thompson. But the one we are talking about is not?
Mr. Wagner. Correct.
Chairman Thompson. Dr. Romine, let's be clear. You
mentioned that African Americans and Asians get misidentified.
Mr. Romine. In the highest-performing algorithms we do not
see that to a statistical level of significance, for one-to-
many algorithms, the identification algorithms.
For the verification algorithms, we do see or the one-to-
one algorithms we do see evidence of demographic effects for
African Americans, for Asians, and others.
Chairman Thompson. Thank you.
The Chair recognizes Ms. Slotkin for 5 minutes.
Ms. Slotkin. Thank you.
Thank you for clarifying that because it was hard to
understand from your testimony.
So just to be clear, Dr.--I am sorry. Can you pronounce
your name? I want to pronounce it right.
Mr. Romine. Romine.
Ms. Slotkin. Romine. Sorry. Apologies.
Mr. Romine. That is quite all right.
Ms. Slotkin. So in a certain segment of these algorithms,
there is some evidence that they have higher rates of mistakes
for African Americans and Asian Americans; is that correct?
Mr. Romine. It is correct that most of the algorithms in
the one-to-many that are submitted do exhibit those
differentials. The highest performing ones in the one-to-many
do not.
Ms. Slotkin. OK. So some do and some do not. I am just
trying to clarify.
Thank you all for being here. I am from Michigan. So we
have a long history of needing our CBP officers to protect us
at the Detroit airport and all of our bridges and crossings. So
can you help me understand?
Is this technology, Mr. Wagner, being used in any way at
our bridge crossings in the Northern Border?
Mr. Wagner. No, not at the bridge crossings.
Ms. Slotkin. OK. But at the airport.
Mr. Wagner. At the airport.
Ms. Slotkin. I know at the airport.
So while I recognize it seems to be a small number of times
or of these programs where they have detected more problems
with particularly African American women I think were mentioned
and Asian Americans, walk me through the process where it would
be you are an average citizen. You are from my district. You
are an African American woman.
Let's say we employ this technology and it shows a
positive, right, a link. Just walk me through that process and
how you would deal with that at the actual border for that
actual citizen.
Mr. Wagner. You would then just show your passport, which
is what you do today, and a person would manually review it if
you did not match.
Ms. Slotkin. If they showed the passport but the technology
still showed a match, what does that officer do in that
situation, if the machine is saying one thing and the passport
is saying another?
Mr. Wagner. We would go on the basis of the document we are
presenting and which photograph we have identified you with or
which identity we have identified you with.
Ms. Slotkin. OK. Then that person would cross the border
and go on with their--I am just asking.
Mr. Wagner. Yes.
Ms. Slotkin. For the average person to understand how this
is being implemented.
Mr. Wagner. What we are matching people against, U.S.
citizens, is that passport photo. We have an electronic copy of
that passport database. So----
Chairman Thompson. Excuse me just a minute.
Staff, you all are being most disrespectful to the hearing.
Please.
Mr. Wagner. So when you are flying into the country, you
preassemble a gallery of those photographs, and that is what we
match you against. So on the officer's screen, they will see
the photograph which should be also what is printed on your
passport, which also should be on that electronic chip in your
passport.
We will look at you and make sure you are all that same
person. If it does not match against that, then we will have to
figure out why.
Ms. Slotkin. When you figure out why, is that individual
allowed to progress?
You know, we got to Windsor to like see a concert, and we
go to Canada quite often in Michigan.
Mr. Wagner. Right. It could be as simple as just looking at
your passport document and saying, ``OK. That's you.''
Ms. Slotkin. OK.
Mr. Wagner. Then we will figure out later what happened.
Ms. Slotkin. Then what happens with that data, right?
So let's say a woman has gone to her concert in Canada.
What happens to her data where it is flagged that she is
falsely flagged that she is matched against someone who has
done something wrong?
What happens in the Department to that information?
Mr. Wagner. If you are a U.S. citizen, the new photograph
we take is discarded after 12 hours. There is no need for us to
keep the new photograph.
There is a record of the transaction that you crossed the
border. If there is some type of error in that, then our
analysts would look at it and correct it basically.
If you have matched, which happens very often in a
biographical sense, your name, date of birth, to the wrong
person even though your biographic match is identical to
someone else, that is where we can also use the facial
recognition to help us distinguish between the people with the
common names.
We can put notes in the system then to advise the officers
to suppress that information from even appearing when we query
your passport the next time.
Ms. Slotkin. Tell me how this technology where you have
been implementing it at different airports and different land
borders, I understand, in the South. Tell me: What are the
results?
How many people have you identified in a positive way that
needed to be identified?
Tell me some statistics to help me to demonstrate the value
of these programs.
Mr. Wagner. Sure. It is 43.7 million people we have run
through it to date at all the different locations, inbound,
outbound, cruise ships, land border pedestrians. We have caught
252 imposters, people with legitimate travel documents
belonging to someone else. I think 75 of those were U.S. travel
documents.
Remember for U.S. travel documents the only biometric we
have is that digitized photo that the State Department has put
on the electronic chip. There is no fingerprint record. There
is no fingerprint requirement to get a U.S. passport.
I am not advocating for one, but there is not one there. So
the only biometric we have on a U.S. travel document is that
digitized photograph, and that is a worldwide standard.
That chip is allowed to be opened by any country
participating in that ICAO scheme that can access the chip and
pull off the digital photograph and then do some type of
comparison to that.
Ms. Slotkin. So in my remaining time, so tens of millions
of people that you have used that have gone through this
technology, just tell me a little bit more about your stats.
How many positive stories? How many negative hits?
Mr. Wagner. So our match rate is about 97 to 98 percent.
That 2 to 3 percent generally means we could not find that
person in that preassembled gallery, meaning we did not match
against anything. We did not match against the wrong person. We
just did not find a match of people traveling.
It could be various environmental or operational reasons
for why that happened.
Ms. Slotkin. How many are false positives?
Mr. Wagner. I am not aware of any.
Ms. Slotkin. OK.
Mr. Wagner. But there may be a small handful. I am just not
aware of any, but as we built this and tested it, we are just
not seeing that.
Ms. Slotkin. I think my time has expired. Thank you,
gentlemen.
Chairman Thompson. Thank you.
The Chair recognizes the gentleman from Texas, Mr. McCaul.
Mr. McCaul. Thank you, Mr. Chairman.
You know, the 9/11 Commission recommended the use of
biometrics for those entering and leaving the United States,
and I believe that technology is our friend in stopping
terrorists and bad actors from entering this country.
We have seen it time and time again, and my first question
is my understanding is the entry/exit program, American
citizens can opt out of that program. Is that correct?
Mr. Wagner. Yes, that is correct.
Mr. McCaul. So there is no requirement that all Americans
have to be subjected to this.
Mr. Wagner. No, but people have to like establish their
identity.
Mr. McCaul. Yes.
Mr. Wagner. Once we determine either through manual review
of the passport or by using the technology, they are a U.S.
citizen and they are excluded from the biometric tracking
requirement.
Mr. McCaul. Right.
Mr. Wagner. But they can opt out of having their picture
taken to make that determination.
Mr. McCaul. It is just like we use with global entry. Most
of my constituents love global entry. You know, I got the CLEAR
Program, as did Mr. Katko, associated with TSA so you could put
your fingerprints down and get to the head of the TSA PreCheck
line.
These are the technologies. I think it made it easier for
the traveling public, but also the great thing is it does not
lie. Biometrics, it is you, and it is hard to fake that.
The last Congress we passed out of the committee my bill,
the Biometric Identification Transnational Migration Alert
Program, otherwise known as BITMAP. Now, I know this is an ICE
program, not CBP, but it passed overwhelmingly in a bipartisan
way on the floor, 272 to 119.
It reauthorizes successful programs started under the Obama
administration that Secretary Jeh Johnson and I talked a great
deal about.
How can we use BITMAP to identify when these people were
coming into our hemisphere?
They may change their names multiple times along the route
to get to the United States, yet their facial recognition,
their biometrics do not. Their names do, but not their
biometrics.
This has been, in my judgment, a very successful program in
keeping terrorists, human traffickers, and bad actors out of
this country.
In fact, this program has enrolled over 155,000 encounters
of persons of interest and 460 known and suspected terrorists,
including arresting violent criminals and rapists involved in
transnational criminal organizations.
So, again, Mr. Wagner, can you comment on why that program
is so valuable to the security of the United States and the
American people?
Mr. Wagner. Sure. It is critically important because, as
you mentioned, people do change their biographic details, and
you know, most of our watch list searches are biographically-
based.
But if we can identify people, especially people traveling
via air, that we have National security concerns about and they
are entering our hemisphere, if they are entering in Central or
South America, we can work with our partners down there and
establish on a biometric basis who that person is so that no
matter what identity they show up in later, if they show up on
the U.S.-Mexico border, we can run the biometric confirmation
to see, well, who were they when they first flew into the
hemisphere. It is critically important.
Mr. McCaul. The travel documents can change and passports
are stolen and manufactured.
Mr. Wagner. Absolutely.
Mr. McCaul. That is not accurate, but the biometrics do not
lie.
Mr. Wagner. Correct. People change documents, steal
documents, borrow documents, purchase documents.
It is harder to alter them now, but the ability to get a
legitimate document that looks like you, and if you can pass by
the visual inspection of somebody glancing at the little 2 2
photograph on it, yes, yes, and that is where the risk is.
Mr. McCaul. It is unfortunate the Senate in its usual
wisdom did not pass this bill. They stole a lot of legislation
the Chairman and I in a bipartisan way passed last Congress,
and that is unfortunate. I would hope we could pass this bill
again this Congress.
I do think we have to look at civil liberties and privacy
as well, but I do think entry/exit is opt-out. It applies
primarily to Americans who would want to opt in and foreign
nationals, and BITMAP applies really almost really to foreign
nationals themselves.
So I want to thank the witnesses for your testimony.
Mr. Chairman, thanks for having this hearing. I yield back.
Chairman Thompson. Thank you very much.
The Chair recognizes the gentlelady from New York, Ms.
Clarke.
Ms. Clarke. Thank you very much, Mr. Chairman. I thank our
Ranking Member. I thank our expert witnesses who testified
before us today.
But it is time that we face the fact. Unregulated, facial
recognition is just not an option. We can debate and disagree
about the exact situations where we should permit the use of
facial recognition, but we should all agree that there is no
situation where facial recognition should be used without
safeguards against bias and protections for privacy.
Right now in terms of regulation, facial recognition is
still in the Wild West. Meanwhile facial recognition
technologies are routinely misidentifying women and people of
color.
Although there are some promising applications for facial
recognition, these benefits do not outweigh the risk of
automating discrimination.
We have seen what happens when technology is widely
deployed before Congress can impose meaningful safeguards. So
let us all look before we leap.
Mr. Wagner, some of our staff have observed issues with
facial recognition technology screening at airports. For
example, we have seen passengers, and particularly darker-
skinned passengers it seems, not able to be matched due to poor
lighting or other factors.
Does CBP track how often its systems failed to capture
photos of sufficient quality for matching?
Mr. Wagner. We track the number of--well, we do not own all
of the cameras. So it is difficult for us to track what an
airline does or how many pictures they might be taking before
they submit one to us for matching. Because in the departure
environment, the airports or the airlines are the ones that own
them.
So we are tracking how many pictures we receive and what
our match rates against them are.
Ms. Clarke. Yes, I was just wondering about the quality
because if the photo quality is not good enough, the accuracy
of the matching algorithm is irrelevant.
Mr. Wagner. Absolutely. So we set a minimum standard. The
picture has to be of this quality before it even----
Ms. Clarke. But do you track the numbers of photos that do
not meet your standard?
You said you have all of these other partners that are
taking photos.
Mr. Wagner. Right.
Ms. Clarke. If you are using their material, then now you
are dealing with something that has become irrelevant if you do
not know what subset of those do not meet your quality control,
right?
Mr. Wagner. Well, we know the pictures that they transmit
to us, whether or not they meet----
Ms. Clarke. Right. But have you----
Mr. Wagner. But we do not know how many attempts they made.
Ms. Clarke. Absolutely, and you do not know the quality.
You do not know how much of that, what percentage of that does
not meet your standard.
Mr. Wagner. Well, we look at the number of passengers on--
--
Ms. Clarke. Do you know the percentage that does not meet
your standard?
Mr. Wagner. Not offhand, no.
Ms. Clarke. OK. How does CBP plan to address these issues
to ensure it can capture high-quality images of travelers for
successful facial recognition screening?
Mr. Wagner. That is the partnership with NIST where we look
at. We have a high-performing algorithm. Now we look at the
operational variables to make that even more high-performing.
Ms. Clarke. So what I would say to you is that then until
you have met that standard, you are not doing the public a
service.
Mr. Wagner. What standard is that?
Ms. Clarke. Of quality control.
Mr. Wagner. We are developing that standard.
Ms. Clarke. Right. It is not developed, right? You're
developing.
Mr. Wagner. Not necessarily, no. I mean, what we are
seeing, if we are matching it at 97 to 98 percent rate, we are
getting----
Ms. Clarke. Let me go on to another question.
When you are in that 3 percent, it does not matter about
the other----
Mr. Wagner. We are not seeing demographic-based, you know,
rates in that 3 percent, and that is when we partnered with
NIST to come in and help us understand that better to be sure
that that is the case.
Ms. Clarke. Very well. Very well.
I understand that since our last hearing CBP completed
operational testing of the biometric entry/exit systems at
airports. The results indicated that the system accurately
matched images when captured, but the rate of successfully
capturing an image was significantly lower than expected, 80
percent compared to 97 percent.
Most of these issues were attributed to airlines reverting
to manually processing passengers to speed the boarding
process. Are you aware of these findings?
Mr. Wagner. Yes, and that is as we were developing the
operational variables to look at, No. 1, does it even work,
right? Can we make it work?
Now we look at and we work with the airlines to not shut
down their boarding. What is the ease of the application of the
traveler engaging with that?
Ms. Clarke. So quickly, what steps is CBP taking to work
with airlines to increase image capture rates?
Mr. Wagner. So one is publishing the regulation, which
would then put the requirement onto the foreign national who
has to comply with the biometric exit Congressional mandate.
Then we can work with the carriers to increase the rate at
which people----
Ms. Clarke. Can you provide our committee with those steps?
That would be helpful.
Mr. Wagner. Sure. Absolutely.
Ms. Clarke. Let me ask just quickly because I have run out
of time.
Mr. Mina, you spoke in your testimony about impermissible
bias, and I was just wondering since you used that terminology,
is there something called permissible bias?
Mr. Mina. I think if I understand your question correctly,
the reason why we used that term ``impermissible bias'' is
because, as Mr. Wagner has talked about and Mr. Romine has
talked about, there are lots of reasons why there may be
failure to cause a match, like, you know, for example,
lighting, environment.
But our office is really focused on an error that is
created based on a protected characteristic, like race or sex
or age. When I make that reference to impermissible bias that
is what I am referring to.
Ms. Clarke. So there is no bias that is permissible. In
other words, if there is a quirk of some sort and you find it
to be so inconsequential that it becomes part of your standard,
that becomes permissible bias?
I am just trying to understand what you mean by
impermissible bias.
Mr. Mina. Again, I think what I am focusing on is what is
actually prohibited by law that our office would look at, which
is really based on those protected characteristics.
Now, of course, you know, obviously CBP and folks across
the Department are trying to eliminate any bias if they find
any reason. However, in terms of what we do as a policy office,
we are really focused on the potential for bias in those
protected areas.
Ms. Clarke. Very well. I yield back. Thank you, Mr.
Chairman.
Chairman Thompson. Thank you.
The Chair recognizes the gentleman from New York, Mr.
Katko.
Mr. Katko. Thank you, Mr. Chairman, and I appreciate you
having this hearing. This is very important.
I commend all of my colleagues for their probing questions
because it is important, but I will make this general
observation based on my time as a prosecutor.
When I was first a prosecutor, DNA evidence was this weird
science thing that no one really knew about, and as we went on
and as it got refined and as it got better, it became a very
potent tool not just for law enforcement, but to exonerate
people who were wrongly accused of crimes.
I see the biometric technology filling a similar role. It
is going to help law enforcement. It is also going to do a
dramatically good thing to prevent misidentification of
criminal conduct, and I am heartened for that.
So one of the things I am heartened most about that I heard
today was from Dr. Romine that the highest-performing
algorithms have no statistical anomalies, if I understand that
correctly.
So that means that at some point those algorithms will get
to the front lines, and I encourage you to get them to the
front lines quickly.
I encourage Mr. Mina never to let your guard down and
always follow any problems with these systems and make it
better because in the end, we are all going to benefit.
I trust my colleagues will ask other probing questions. So
I have to ask something of Mr. Wagner that occurred yesterday
that is very important to my constituents, in general, but to
New York State, in particular.
A letter was sent February 5, 2020, which was yesterday, to
New York State saying why Homeland Security can no longer have
New York driver's licenses as part of the formula for the
Trusted Traveler Program, and that is because New York State
under the Green Light Law, which it passed, forbids access by
CBP and ICE to the New York driver databases.
So could you briefly summarize for us, and I will ask that
this letter be incorporated into the record, Mr. Chairman,
first of all.
I ask that the letter be incorporated into the record.
Chairman Thompson. Without objection.
[The information referred to follows:]
Letter From Chad F. Wolf, Acting Secretary, U.S. Department of Homeland
Security
February 5, 2020.
Mark J.F. Schroeder,
Acting Commissioner, New York State Department of Motor Vehicles, 6
Empire State Plaza, Albany, NY 12228,
[email protected].
Theresa L. Egan,
Executive Deputy Commissioner, New York State Department of Motor
Vehicles, 6 Empire State Plaza, Albany, NY 12228,
[email protected].
Via email and U.S. mail
Dear Mr. Schroeder and Mrs. Egan: On June 17, 2019, the State of
New York (New York) enacted the Driver's License Access and Privacy Act
(the Act), effective December 14, 2019.\1\ The Act forbids New York
Department of Motor Vehicles (DMV) officials from providing, with very
limited exceptions, pertinent driver's license and vehicle registration
information to the United States Department of Homeland Security (DHS).
Specifically, this Act precludes U.S. Customs and Border Protection
(CBP) and Immigration and Customs Enforcement (ICE) from accessing and
validating pertinent information contained in New York DMV records that
is operationally critical in DHS's efforts to keep our Nation secure.
The Act also threatens to block access to other State law enforcement
agencies and departments if those agencies or departments provide New
York OMV records to CBP and ICE.
---------------------------------------------------------------------------
\1\ N.Y. Veh. & Traf. 201 (2019).
---------------------------------------------------------------------------
Over the years, CBP has utilized New York DMV records in several
ways to promote national security and to enforce Federal customs and
immigration laws. Having access to New York DMV information has enabled
CBP to validate that an individual applying for Trusted Traveler
Programs (TTP) membership qualifies for low-risk status or meets other
program requirements. An individual's criminal history affects their
eligibility for TTP membership. TTP permits expedited processing into
the United States from: International destinations (under Global
Entry); Canada only (under NEXUS); and Canada and Mexico only (under
SENTRI). TTP also allows quicker processing for commercial truck
drivers entering or exiting the United States (under FAST).
Furthermore, CBP has needed New York DMV records to establish ownership
and thus to determine whether a used vehicle is approved for export.
The Act prevents DHS from accessing relevant information that only
New York DMV maintains, including some aspects of an individual's
criminal history. As such, the Act compromises CBP's ability to confirm
whether an individual applying for TTP membership meets program
eligibility requirements. Moreover, the Act delays a used vehicle
owner's ability to obtain CBP authorization for exporting their
vehicle.
Furthermore, on a daily basis, ICE has used New York DMV data in
its efforts to combat transnational gangs, narcotics smuggling, human
smuggling and trafficking, trafficking of weapons and other contraband,
child exploitation, exportation of sensitive technology, fraud, and
identity theft. In New York alone, last year ICE arrested 149 child
predators, identified or rescued 105 victims of exploitation and human
trafficking, arrested 230 gang members, and seized 6,487 pounds of
illegal narcotics, including fentanyl and opioids.\2\ In the vast
majority of these cases, ICE relied on New York DMV records to fulfill
its mission. ICE also needs New York DMV information to safeguard
Americans' financial and intellectual property rights.
---------------------------------------------------------------------------
\2\ Nation-wide, last year ICE arrested nearly 4,000 child
predators, identified or rescued 1,400 victims of exploitation and
trafficking, arrested 3,800 gang members, and seized 633,000 pounds of
contraband, including fentanyl and opioids.
---------------------------------------------------------------------------
New York DMV records have long been used by ICE law enforcement
personnel to verify or corroborate an investigatory target's Personally
Identifiable Information (PII), which can include their residential
address, date of birth, height, weight, eye color, hair color, facial
photograph, license plate, and vehicle registration information.
Moreover, ICE's expeditious retrieval of vehicle and driver's license
and identification information has helped identify targets, witnesses,
victims, and assets. ICE has used DMV records to obtain search
warrants, and DMV records are also critical for ICE to identify
criminal networks, create new leads for investigation, and compile
photographic line-ups. Additionally, during the execution of search-
and-arrest warrants, ICE officers have used DMV information to identify
individuals whose criminal history renders them a threat. The Act
prohibits the sharing of vehicle registration information, including
the identity of the person to whom the vehicle is registered, with DHS.
That prohibition prevents ICE from running license plate searches, even
when ICE is aware that the vehicle's owner has committed a heinous
crime. In short, this Act will impede ICE's objective of protecting the
people of New York from menacing threats to national security and
public safety.
Although DHS would prefer to continue our long-standing cooperative
relationship with New York on a variety of these critical homeland
security initiatives, this Act and the corresponding lack of security
cooperation from the New York DMV requires DHS to take immediate action
to ensure DHS's efforts to protect the Homeland are not compromised.
Due to the Act's negative impact on Department operations, DHS will
immediately take the following actions:
(1) Trusted Traveler Programs--Global Entry, NEXUS, SENTRI, and
FAST.--Because the Act prevents DHS from accessing New York DMV
records in order to determine whether a TTP applicant or re-
applicant meets program eligibility requirements, New York
residents will no longer be eligible to enroll or re-enroll in
CBP's Trusted Traveler Programs.
(2) Vehicle Exports.--Because the Act hinders DHS from validating
documents used to establish vehicle ownership, the exporting of
used vehicles titled and registered in New York will be
significantly delayed and could also be costlier.
These actions are the result of an initial assessment conducted by
DHS. We will continue to review Department-wide operations related to
New York to assess and mitigate the Act's adverse impact on national
security and law enforcement.
Sincerely,
Chad F. Wolf,
Acting Secretary.
Mr. Katko. Thank you.
Could you just briefly summarize the contents of this
letter? Then I have a follow-up question for you.
Mr. Wagner. So my understanding is New York State because
of the law that they passed, you know, without consultation
shut off the access to motor vehicle data, which included
driver's license information, license plate registration,
vehicle registration information.
So in our operations, any of the work that we do where we
would use that information to help validate an identity, an
address, and a vehicle, the ownership of a vehicle is impacted
by not being able to do that directly, and the breadth of our
mission goes way beyond, I think, what the law says about
immigration enforcement.
You are impacting the Customs mission and the National
security mission, and all the other areas in which we operate.
Mr. Katko. Is there any other State in the country that is
having this problem with Customs?
Mr. Wagner. We have worked some other agreements with other
States to continue to access the data for, you know, the work
that we do.
Mr. Katko. So am I to understand that New York State is the
only one who forbids Customs and Border Protection as well as
ICE to have access to their driver databases?
Mr. Wagner. Yes. It is the only one I am familiar with
right now.
Mr. Katko. Even California?
Mr. Wagner. California, we have a separate agreement with
where we continue to access their information.
Mr. Katko. Now, I just want to note further as long as we
have a couple of minutes here some of the things that are in
the letter. Tell me if this is correct.
On a daily basis, ICE uses New York DMV data in an effort
to combat transnational gangs, narcotics smuggling, human
smuggling and trafficking, trafficking of weapons and other
contraband, child exploitation--child exploitation?--
exploitation of sensitive technology, fraud, and identity
theft.
Is it fair to say that by not having access to the
database, it hampers those investigations at times?
Mr. Wagner. Sure. Any law enforcement practice where you
would normally use that information, yes, would be impacted.
Mr. Katko. I yield back the balance of my time.
Chairman Thompson. Thank you.
The Chair recognizes the gentlelady from New York, Miss
Rice, for 5 minutes.
Miss Rice. Thank you, Mr. Chairman.
Let's continue, Mr. Wagner, if we can, talking about what
happened with New York.
Was CBP made aware of the policy before the Acting
Secretary's announcement on Fox News?
Mr. Wagner. Yes.
Miss Rice. So you were aware of it. No notification was
made to Congress about blocking access to these Federal
programs for New Yorkers?
Mr. Wagner. I do not know.
Miss Rice. Well, there was none.
So personally, we, my office has already received an influx
of new questions about this policy literally overnight. Fifty
to 80,000 New York State residents are affected who have
pending global entry enrollment applications or renewals.
This is going to have an enormous impact on people, many of
whom entered into this program because their jobs require them
to travel internationally.
So what do you plan to do about all those people who are
going to be impacted?
Mr. Wagner. Well, without the ability to help validate
their identity through the----
Miss Rice. You have their fingerprints.
Mr. Wagner. Yes, but if they have not been arrested, the
fingerprints do not tell us anything. What would the
fingerprints tell you if you have not been arrested?
Miss Rice. So what are you trying to find out is my point.
Mr. Wagner. Trying to validate their address where they
live, their residency. These are things important to us as we
establish that low-risk Trusted Traveler status that we afford
people in that program. Without the ability to do that, how
would we do that?
So New York State shut off without consultation our access
to that information in December. How would we continue to
operate and validate who people are?
Miss Rice. Well, going forward, what about the people who
already have it?
I have global entry. So when I go to renew it, I am not
going to be able to do that.
Mr. Wagner. Correct.
Miss Rice. Yet here I am, a sitting Congresswoman with
global entry. So to me, to me, to me, I understand the
distinction that you are making. There are at least 15 other
States you are saying that you have individual agreements with
all of them where they do not block access to this database?
Fifteen other States who have a global----
Mr. Wagner. I am not aware of any other State blocking our
access to that information.
Miss Rice. OK. So I would like you--we are going to follow
up. I am going to follow up directly with you because there are
at least 15 other States that allow undocumented people to get
driver's licenses.
Mr. Wagner. Yes.
Miss Rice. I would----
Mr. Wagner. I am not aware of them blocking our
information.
Miss Rice. OK. So you not being aware is not a sufficient
answer because there could be other States that do, and it
seems to me that this is, once again, an attempt by this
administration, specifically Donald Trump, who formerly was a
New Yorker, to punish New York.
So you and I are going to follow up on this, and I
appreciate you trying to answer these questions, but we need
more information, and I appreciate your attempt to answer these
questions.
I yield back. Thank you, Mr. Chairman.
Chairman Thompson. Thank you very much.
Mr. Wagner, just for the record, can a person have global
entry without a driver's license?
Mr. Wagner. Yes, I believe so.
Chairman Thompson. So I am trying to figure out how you are
going to cancel all of these people and some of them do not
even drive and deny them.
Mr. Wagner. Well, it is a New York State identification.
Chairman Thompson. But they have passports.
Mr. Wagner. Validation of that information.
Chairman Thompson. But they have a passport. They have a
passport.
Mr. Wagner. How do we validate the address of where they
live?
Chairman Thompson. My driver's license has a post office
box. So, I mean, I am just trying to figure out are you being--
--
Mr. Wagner. Why is the information blocked for this purpose
then?
Chairman Thompson. Well, I do not know. I am saying why
would you cancel it if it----
Mr. Wagner. Well, why would New York State block the
information for this purpose?
Chairman Thompson. Is it for identification or security?
Mr. Wagner. Both.
Chairman Thompson. But you can prove it with other
documents. I mean, that is what I am trying to figure.
Well, the Chair recognizes the gentleman from Louisiana,
Mr. Higgins.
Mr. Higgins. Thank you, Mr. Chairman.
I yield 1 minute to my colleague, Mr. Katko.
Mr. Katko. Thank you, Mr. Higgins.
Just have a quick follow-up question with my colleague, and
it is a quite simple one really.
First of all, it is clear that it hampers investigations
with ICE. It is clear that it hampers the ability to get
certain identification that is available in driver's DMV
database in New York State.
I just want to make sure that it is clear. My colleague
from New York, Miss Rice, mentioned that there are many other
States that have possible--like allowing illegal aliens to get
driver's licenses.
That is not the issue. The issue is, is there any other
State in the United States of America that completely blocks
Customs and Border Protection and ICE's access to DMV records.
Mr. Wagner. I do not believe so.
Mr. Katko. OK. So in my opinion, and I have an immense
amount of respect for my colleague from New York, I do not
believe this is a political exercise. All New York would have
to do is enter into a similar agreement that those other 15
States have entered into with Customs and Border Protection and
ICE where they simply verify that they will not use it for
immigration enforcement purposes, but use it for law
enforcement purposes and for global entry and those types of
things.
Is that correct? You can do that?
Mr. Wagner. I think that is a discussion we would have with
the State.
Mr. Katko. OK. You have done it with other States?
Mr. Wagner. Yes.
Mr. Katko. OK. Thank you.
I yield back.
Mr. Higgins. I thank my colleague.
Just to follow up on the New York question because it is
just a fascinating topic, are you aware of negotiations or
communications prior to the New York legislative body passing
this law with Customs and Border Protection?
Were we out front with this communication at all?
Mr. Wagner. Our access----
Mr. Higgins. It seems to me like they should have known
before they passed the law this was going to happen.
Mr. Wagner. Right. Our access was just turned off one day
in December, and our officers and agents in the field called in
and said, you know, ``What happened to our access?''
Mr. Higgins. So you are saying that as far as you know, and
you can certainly advise if you do not know or have no way of
knowing, but as far as you know, sir, was there an on-going
communications during the course of the development of this
legislation in the State of New York with the law enforcement
agencies like Customs and Border Protection and ICE?
Mr. Wagner. I do not know. I am not aware of any.
Mr. Higgins. Well, one would hope that there was.
Dr. Romine, you mentioned black box texting. Would you
clarify that that means that as your products are tested
through NIST, your facial recognition products provided by
vendors, that they are tested without your knowledge of who the
vendor is? You are strictly looking at the results of the
algorithms themselves?
Mr. Romine. So when I use the phrase ``black box testing,''
what I mean is that we do not have any insight into the
characteristics of the algorithm itself. We publish an API, an
application----
Mr. Higgins. Do you know the identity of the vendor?
Mr. Romine. It is self-identified.
Mr. Higgins. It is self-identified as you are studying the
product itself.
Mr. Romine. That is correct. We do that.
Mr. Higgins. OK. Just to clarify that.
Now, can any vendor submit an algorithm to NIST for
testing?
Mr. Romine. Yes, sir.
Mr. Higgins. The process by submitting that product is
standardized?
Mr. Romine. It is, sir.
Mr. Higgins. All right. With the top-performing algorithms
like Customs and Border Protection uses, is there a wide
variance between what you are referring to as the top-
performing algorithms and, say, academic projects perhaps
submitted for testing?
Mr. Romine. Yes, sir. There is a wide variance in the
performance of algorithms at the top.
Mr. Higgins. So in your scientific assessment of NIST
testing and evaluation of facial recognition technologies,
would you say that what we are referring to as the top-
performing algorithms that are being used by Customs and Border
Protection are far and beyond some of the common products that
are presented to you?
Mr. Romine. The top-performing algorithms are significantly
better in their error rate.
Mr. Higgins. Can you confirm for this committee, sir, that
it is, indeed, the top-performing algorithms at this point that
are being used by Federal law enforcement agencies?
Mr. Romine. Sir, I have no way to independently verify
that.
Mr. Higgins. But would you say that Customs and Border
Protection is using the top. I want to confirm.
Mr. Romine. I did not say that.
Mr. Higgins. Can you confirm that, good sir?
Mr. Romine. We are using not the algorithm they tested, but
we are using the previous version of it, and we are switching
to, we are upgrading to the version that they tested next
month.
So, yes, we are using a high-performing vendor.
Mr. Higgins. Going to the next iPhone? All right. I think
that vaguely answers my question, and my time has expired.
Mr. Chairman, thank you.
Chairman Thompson. Thank you very much.
The Chair recognizes the gentleman from New Jersey, Mr.
Payne.
Mr. Payne. Thank you, Mr. Chairman.
Let me ask. Who and where is all of this facial recognition
data stored?
Please describe under what specific circumstances this data
is allowed to be shared or used or transferred, if that is the
case?
Mr. Wagner. We are using as a database travel document
databases. So these are photographs collected by the U.S.
Government for the purposes of putting on a travel document,
like a U.S. passport or a U.S. visa that is issued to a foreign
national or a photograph of a foreign national when they arrive
in the United States, like under the Visa Waiver program.
We would take their photograph or read the photograph from
the chip in their passport and store that. That is what forms
the baseline gallery that we match against.
Now, new photographs we take of a person, U.S. citizen, if
we match it to a U.S. passport or a U.S. identity, those photos
are discarded, OK, after 12 hours just for some system work.
If you are a foreign national, that goes over to a system
called IDENT that DHS runs where they are stored under the
protocols of the Systems of Record notice of the data retention
period of that, which I believe is 75 years.
Mr. Payne. OK. All right. To follow up with that, you know,
we are living in an age where everything is being hacked. What
type of security measures or protections have been put in place
regarding the security of this data?
Mr. Wagner. So the databases are housed within the U.S.
Government. CBP does not necessarily keep or own any of those
permanent databases. You know, they are owned by Department of
State. They are owned by other branches of DHS.
We access a lot of that information. We use it. We match
against it, and then we put information back into them.
Mr. Payne. OK. You know, I continue to have hits come
across my desk about the mishaps and disadvantages of facial
recognition technology and the racial bias. It is my
understanding that the technology continues to misrepresent and
irregularly identify people of color and women.
So am I hearing from the majority of the panel that that is
not the case? Because it keeps coming to us. So there has to be
some validity.
Mr. Romine. Sir, in our testing for the one-to-one
identification algorithms, we do see evidence of demographic
effects, differences with regard to race and sex and age.
Mr. Payne. OK.
Mr. Romine. In the one-to-many identification testing that
we did for the algorithms that we tested, there was a small set
of high-performing algorithms that had undetectable
differentials.
Mr. Payne. OK.
Mr. Romine. But the majority of the algorithms still
exhibit those characteristics.
Mr. Payne. Can you give a description of the difference
between the two sets?
Mr. Romine. Yes, sir. In the case of verification,
verifying an identity, a biometric is matched solely or in the
case of face recognition a picture is matched against----
Mr. Payne. Is that the one-to-one?
Mr. Romine. That is the one-to-one.
Mr. Payne. All right.
Mr. Romine. The identification or the verification is to
try to determine if you are who you say you are.
Mr. Payne. All right.
Mr. Romine. It is matched against a gallery of one, in
essence.
Mr. Payne. What is the one-to-many?
Mr. Romine. The one-to-many is matched, in the case of
CBP's application, one to perhaps thousands for the airline
public, the traveling public, or one to millions in the case of
law enforcement, such as FBI, to try to identify a suspect.
Mr. Payne. So you are saying that the percentage of
identifications in the one-to-one, you have more incidents of
this bias that we see?
Mr. Romine. I should clarify. In the algorithms that we
tested, that is correct. However, many of the vendors who chose
to participate in the one-to-many testing did not choose to
participate in the one-to-one, and those are some of the
highest-performing in the one-to-many.
Mr. Payne. OK. Thank you.
I yield back, Chairman.
Chairman Thompson. Thank you.
The Chair recognizes the gentleman from North Carolina, Mr.
Walker.
Mr. Walker. Thank you, Mr. Chairman.
I would like to yield 1 minute to the gentleman from
Louisiana.
Mr. Higgins. Thank you, my colleague.
Dr. Romine, I have a question regarding the effectiveness
of the technology that you have tested regarding children.
Is it a potential if we assembled a gallery of photographs
of children crossing the border, some of whom are being
exploited and false identifications presented; how does the
technology work with children compared to mistake and errors in
other demographics?
Can this technology be used to protect children that are
perhaps being exploited crossing our borders, coming into our
country?
If so, what can we do to protect the privacy of those
children, given the fact that they are minors?
I will leave you my remaining 30 seconds here.
Mr. Romine. Thank you, sir.
The application specifically is something that we do not
test. What we have tested is the effectiveness of the
algorithms in terms of error rates.
We do find that for children in the one-to-one setting, the
one that you just described, there are demographic effects
there. There are differentials. The error rates are higher in
the one-to-one case with respect to age.
So it is more difficult. Based on our testing, it appears
more difficult to match.
Mr. Higgins. But there is no gallery. There is no one-to-
many. There is no gallery of the photographs that you have.
Mr. Romine. We have no such gallery.
Mr. Higgins. If we did develop that, then NIST could test
the effectiveness and perhaps this could be a tool to protect
children?
Mr. Romine. We could. We could undertake many different
kinds of testing to determine the effectiveness of those.
Mr. Higgins. Thank you, sir.
I thank my colleague for yielding.
Mr. Walker. Absolutely. Thank you, Representative Higgins.
Mr. Wagner, is it true that a biometry entry/exit system
uses less personally identifiable information than the current
system that we have in place?
Mr. Wagner. Yes, because currently you open your passport
booklet and show it to an individual to either, say, check your
bags, go through TSA screening, board the plane, a CBP officer.
You are exposing your name, your date of birth, your passport
number, your place of birth, all the information on your
passport page.
Somebody could be looking over your shoulder. Somebody
could take a picture over your shoulder looking at that. You
are disclosing it to a person who does not actually need to
know all of that additional information versus standing in
front of a camera with no identifiable information other than
your face, which they can already see, and your picture is
taken and on the screen comes a green checkmark, and that
person now knows you have been validated by the Government
record to proceed.
So you are sharing actually less information in this
instance.
Mr. Walker. But not only sharing less information, but on a
scale of 1 to 10, 10 being the highest, how would you rate this
progress as, in your own words, continuing to develop and
rightfully so, would be the highest security possible for
travelers compared to anything else that we are doing now?
Mr. Wagner. Right now I think on top of everything else we
are doing, it brings us closest to 10, which is where we want
to be.
Mr. Walker. When Representative Higgins talked about some
of the children involved, are there any numbers or statistics
based on people that you have caught either involved in human
trafficking or some other nefarious activity because, strictly
because of the facial recognition?
Mr. Wagner. Yes. On the land border, we have got 247
imposters so far, meaning they had a legitimate document that
belonged to somebody else. Eighteen of those, so 7 percent,
were under the age of 18. So they would be considered children.
Seventy-three of those at the land border had U.S.
passports or U.S. passport cards, and 46 of them, or almost 20
percent, had criminal records that they were trying to hide.
Mr. Walker. Do you believe these were identified strictly
because of the use of facial recognition or was there any other
aspect or involvement?
Mr. Wagner. Our officers are also very good at identifying
the behaviors in the person when they present the travel
document. A lot of times that can also be a cue that the
person's hiding something.
But the technology on top of officer's skills and abilities
should bring us to that security posture that will bring us to
near perfect.
Mr. Walker. Are there any policy's difference between a
U.S. citizen versus non-citizen?
Mr. Wagner. Well, everyone has to establish their identity
by law. Everyone has to produce some type of identification.
The law requires a U.S. citizen to travel on a passport.
Mr. Walker. In the process, scrubbing this 12 hours after
is what?
Mr. Wagner. That is our internal policy. We take a new
picture. We discard it after 12 hours. We are looking at
actually shrinking that to a less time. We only keep it there
in case the system crashes and we have got to restore
everything.
Mr. Walker. Thank you, Mr. Wagner.
I yield back, Mr. Chairman.
Chairman Thompson. Thank you.
The Chair recognizes the gentlelady from Las Vegas, Ms.
Titus.
Ms. Titus. Thank you, Mr. Chairman.
I find this interesting. The more you talk the less I know,
it turns out, unfortunately.
McCarran Airport is in my district. It is a very busy
airport, one of the busiest in the country. A lot of
international tourists come through there.
So I know we have talked a lot about the use of this facial
recognition for security reasons. I would like to talk about it
in terms of how it affects the passengers' experience. We want
people in Las Vegas to have a good experience from the time
they land until the time they leave.
So how do you work to coordinate using this for security
and also reducing wait times or serving the passenger as
opposed to making it more difficult for the passengers?
Mr. Wagner.
Mr. Wagner. So it absolutely supports our travel and
tourism goals as well. It makes a much better passenger
experience, a more convenient passenger experience, a more
consistent passenger experience.
You think as you go through the airport the number of stops
you have to make to produce a piece of paper or open your
passport again or provide some other form of validation to go
forward.
You can use the facial recognition and the camera to have
that same process. It is quick enough that you walk up and your
picture is taken and 2 to 3 seconds you are moving forward.
So what we are seeing is reduced wait times. The airlines
as they incorporate it into the boarding process are reducing
their boarding times over the aircraft sometimes as much as,
say, 40, 45 percent.
It is a different atmosphere for the travel because you are
not fumbling for your documents or forgetting where you put
your boarding pass or getting stuck in line behind the person
whose phone went dead when they went through to read their
boarding pass or forgot where they put their passport.
So it is creating a better atmosphere for the traveler. It
is moving the lines quicker because you cannot leave your face
on the plane. You cannot, you know, leave your face in the
bathroom. You cannot forget that like people do with their
travel documents.
So it is making an easier process because everybody knows
how to take a picture, and what we see is people are enjoying
this process a lot better for them, and what we are seeing is
the lines reduced.
Ms. Titus. Are you working with TSA or local law
enforcement to make this all run smoothly or is that not
necessary?
Mr. Wagner. So we are working very closely with TSA. We
have run a few pilots with them. We have an on-going pilot in
Atlanta because we build the gallery as the person prints out
their boarding pass. So anyplace now where they have to show
their passport at the airport, say, when they are departing the
United States, you could take a picture and validate it against
our gallery.
So you are outside of the airport or you just walked into
the airport. You got your boarding pass. Your picture goes into
that gallery.
So steps like checking your bags where you have to show
your ID to the airline person, you can have a camera there that
does that.
You go up to the TSA checkpoint. TSA can take a photograph.
It transmits to our gallery, again, because we built it for the
biometric exit requirement, but we want to make that
environment available to all the other places in the airport
where you would show your passport to do that.
So, yes, so for TSA you could take a picture. Then you go
through screening. You go to board the plane. The airline takes
your picture. It comes back to our gallery. We confirm it. You
can board the plane without even showing your passport to the
airline or showing your boarding pass to the airline.
Ms. Titus. Well, suppose you find somebody does not match.
My understanding is this goes through an app, and law
enforcement, if they are busy or if the person responsible for
checking out the non-match is doing something else, do you have
some kind of staffing model for who is responsible for that?
Because I had heard it comes through an app, and since
there is no action that you are supposed to take that is very
clear, sometimes they just ignore it.
Mr. Wagner. Depending on where you are in the airport,
generally it would be the airline or CBP, we would just look at
the physical passport, which is what you are presenting now,
and we would make a determination.
Now, if we have doubts about do you match the picture on
your passport, right, which happens, or if the airline has
doubts you match the picture on your passport, they may call us
over. We may ask the person for another form of ID. We may ask
them additional questions. We may do a further inspection on
them.
So if you do not look like your passport photo, you know,
from a visual review, these are the same kind of things that
would occur.
Ms. Titus. We have been having a lot of confusion about
going to the Real ID from just regular driver's licenses.
People do not know they have to do that. We are trying to get
the word out.
Some States did not provide the funding to go to Real ID.
Is that transition part of your consideration as you develop
this new system or is it not connected?
Mr. Wagner. It is separate than this.
Ms. Titus. So that it is not going to make any difference?
Mr. Wagner. Not really.
Ms. Titus. All right. So people who will use their
passports instead of Real ID, that will not matter?
Mr. Wagner. Right, because these are international
travelers we are talking about today. So they would generally
have a passport.
Ms. Titus. You do not see this moving to national as well
as international once it is up and running?
Mr. Wagner. I would defer to TSA on that for their
requirements on how this might apply to a domestic flight.
Ms. Titus. OK.
Mr. Wagner. I think there is some good discussion to have
there, that if people have passports and you could
electronically confirm them, even on a domestic flight, should
the traveler opt into this, I think it would be good government
to build a system like this if that is what people would want.
Ms. Titus. Well, thank you.
Thank you, Mr. Chairman.
Chairman Thompson. Thank you.
The Chair recognizes the gentlelady from Illinois, or I am
sorry. The gentleman from Texas, Mr. Crenshaw.
Mr. Crenshaw. Thank you, Mr. Chairman.
Thank you all for being here. It has been an interesting
hearing to watch.
I just want to dispel any misinformation on the facial
recognition technology that we are discussing here today. It
seems to be abnormally controversial.
We are not talking about 1984-style Government
surveillance, not like China has. We are not talking about
facial recognition at the National Mall or Times Square or
downtown Houston.
We are talking about facial recognition at air, land, and
sea ports of entry, where the Government has not just the
authority but the duty, the responsibility to know who enters
our country and where they are already checking for
identification, of course.
It seems from the answers we have gotten that CBP is using
the best algorithms with almost no bias whatsoever in them.
That is what we have established today as far as I understand.
Locations where facial recognition technology is employed,
those locations are marked, correct?
Mr. Wagner. Yes. It is where you would normally present
your passport.
Mr. Crenshaw. OK. Locations where facial recognition
technology is employed where entrants are required to present,
and you already answered that one, present a form of
photographic identification already.
Entrants are allowed to opt out of facial recognition
technology and present photographic identification to a CBP
officer who will then compare the physical appearance of the
entrant and the photographic identification presented, correct?
Mr. Wagner. Correct.
Mr. Crenshaw. Biometric data for U.S. persons is stored for
no more than 12 hours in an encrypted virtual private cloud,
correct?
Mr. Wagner. Correct.
Mr. Crenshaw. Biometric data for entrants who are not U.S.
persons are stored in IDENT, correct?
Mr. Wagner. Correct.
Mr. Crenshaw. Giving the above and knowing where facial
recognition technology is used, requirement to present photo
ID, the ability to opt out, and a secured storage, what are the
major privacy concerns I might be missing?
How can we improve this?
Mr. Wagner. I think what we have heard from the privacy
community is people get used to the convenience of this
technology and that bleeds over into the commercial world or
their private sense, and they may be more likely to allow that
to happen outside of the Government requirements.
You know, in my discussions with them, I said, ``Yes, but
there is also an expectation by the public that they have this
convenience in their private life and why should their
interactions with their Government be so antiquated?''
Mr. Crenshaw. Yes.
Mr. Wagner. Why should their travel through the airport be
so antiquated and manual and frustrating?
You know, do they not expect that that same convenience
should apply when they are traveling internationally?
Mr. Crenshaw. One way this could be viewed in a very
positive sense is to combat human trafficking. Is there a way
that tools like this can be integrated with other tools like
Spotlight and SAFER to battle child sex trafficking, human
trafficking?
Mr. Wagner. Sure. Because what this helps us do is our core
vetting processes are biographically-based, right? A name and
date of birth is submitted, say, to a watch list, you know,
through an airline application, through TSA, and we vet and do
those background checks on the basis of, say, who the airlines
tell us who is flying, so who checked in, who purchased a
ticket.
But when you can then use a biometric to validate that you
vetted the right person, you have the assurances that that is
the person who is actually traveling and not just their
passport is traveling under a different person that is being
trafficked.
So it helps us close those vulnerabilities of imposters for
nefarious or being trafficked or being victimized to be able to
do that using imposter documents.
Mr. Crenshaw. In my limited time left, can this be used to
combat visa overstays as well?
Mr. Wagner. Yes. Now, we track visa overstays primarily
through the biographic information the airline provides, but by
implementing this system, we have actually biometrically
confirmed almost 44,000 by overstays. With the biometric
validation that these people overstayed, they end up leaving
the United States, albeit late, later than they were authorized
to do so. So, you know, just about 44,000.
Mr. Crenshaw. Thank you. I yield back.
Chairman Thompson. Thank you.
The Chair recognizes the gentleman from New York, Mr. Rose.
Mr. Rose. Thank you, Mr. Chairman.
Mr. Mina, the NYPD has in the past used facial recognition
to compare photos from crime scenes against its own internal
arrest databases. Some State lawmakers want to take that
ability away from the NYPD and other New York State law
enforcement agencies.
Do you support police agencies using facial recognition in
the course of their criminal investigations?
Mr. Mina. Congressman, that is not necessarily an issue
that we have looked at at CRCL. We are primarily looking at the
DHS uses of facial recognition technology and, in particular,
we have focused primarily, as Dr. Romine mentioned, less so on
the identification piece where you have sort-of you are trying
to match a photo to a gallery of, you know, tons, and we are
looking at a much narrower. We are looking at, I think, more of
the verification, if I understand the technology correctly.
Again, our role there is really to make sure that we are
addressing these concerns regarding impermissible bias, whether
that is, again, based on race, national origin, age, gender, as
we have talked about.
Mr. Rose. So one thing that I think has been absent in this
conversation is the ways in which civil liberties can
potentially be infringed upon in the absence of the use of
technology. Can you speak to this for a minute or two?
I am thinking of false positives. I am thinking of people
who are being arrested or at least questioned further based off
of just a verbal description.
Mr. Mina. Absolutely, Congressman. So I think that it is
obviously critically important to blend both the use of
technology as well as the end-user in this process.
I do not think it is an either/or proposition, and as we
have advised CBP and other DHS components, that is, from a
policymaking perspective, that is really where we see the
greatest benefit, is really that interaction between the
technology and the user.
Because, as Mr. Wagner talked about earlier, for example,
if there was a false negative, for example, then you would have
the line officer looking or agent looking at their actual, you
know, passport or other travel documentation and making that
independent verification.
Mr. Rose. Sure.
Mr. Mina. Then if it matches, the person goes along and
they board the flight.
Mr. Rose. Right. I think that it is important to know then
so that we are all on the same page that the use of technology
has consistently been implemented to preserve our public
safety, but also to further protect civil liberties.
This is being lost in this conversation as yet again I
think we are unnecessarily politicizing an effort to keep us
safe.
It is not perfect, and you all have some work to do to make
it even better, and I am encouraged to hear that you are making
it better.
Mr. Wagner, you are going to have to hear from another New
Yorker. So look. I am not a supporter of this New York
legislation that was passed. I think it is unfortunate and
wrong that you all were not notified, but two wrongs do not
make a right.
So I am going to ask some very simple questions. If you all
were setting out to be the professional force that you are and
do this professionally, do you think that in advance of
announcing this you should have told Congress what was wrong
and what would happen if it was not fixed or addressed?
Mr. Wagner. I would have to defer to DHS on that.
Mr. Rose. No. Come on, man. This is ridiculous. It is a
simple question. That is a simple question.
We heard about this from Fox News. This is politics at its
worst. We are talking about acting like professionals right
now.
If there is a problem that needs to be addressed and you
all are doing this, do you think it was appropriate that we
were not told well in advance so we can try to arrive at some
solution?
Do you think that is OK? Is that the way you would want to
carry this out?
Mr. Wagner. I am not going to comment on that. I mean, that
is your----
Mr. Rose. You are not going to comment. So by the fact that
you have given very clear and declarative answers previously, I
think that we can all assume what you are thinking and
unwilling to say right now.
So let's commit to actually trying to solve problems here.
You have got Members of Congress that will not be able to renew
something. You have got more important than Members of
Congress. Who cares about Members of Congress? Millions of
other people that are now held in the balance, people on my
staff, people, colleagues, all types of people, all types of
people.
This is politics. If you really were making an effort to
address a problem, to address a problem, there would have been
a system, a proposal, a negotiation, a conversation, letters
written. That is the way business is conducted.
So let's put that aside. Would you now commit, now that we
have all engaged in our politics, to actually having sensible
meetings and conversations about a way forward to solve this
issue?
Mr. Wagner. Sure, I think that is a good point.
Mr. Rose. You would commit to that. OK. Thank you.
Chairman Thompson. The gentleman's time has expired.
The Chair recognizes the gentlelady from Illinois, Ms.
Underwood.
Ms. Underwood. Thank you, Mr. Chairman.
Many of my constituents in Northern Illinois have to drive
over an hour to get to a major airport in Chicago and,
therefore, we are always interested in learning more about
technologies that can improve airport security wait times, but
biometric data is ripe for potential abuse and misuse, which is
why it is so important to ensure that DHS uses facial
recognition and other technologies in a fair and reliable and
effective way.
Mr. Wagner, although children under the age of 14 are not
required to be screened, many do go through screening that
collects their biometric information.
How does CBP store and secure this information? I am
talking about under 14.
Mr. Wagner. I think if you are outside the scope of the
biometric tracking requirement, which is 14 to 79, I believe we
discard all of that information. Let me verify that.
Ms. Underwood. Yes. Would you be willing to provide the
committee with that information in writing?
Mr. Wagner. Yes.
Ms. Underwood. Both the procedure and the policy in order
to do so?
Mr. Wagner. Yes.
Ms. Underwood. OK. Are there any differences in how CBP
collects, uses, or secures children's biometric information in
comparison to adults?
So if a child presents, does it take it and immediately
release, right, or is it going to be going through some kind of
filtering later on?
We want that level of information.
Mr. Wagner. OK. You have got it.
Ms. Underwood. OK. The December 2019 NIST report found that
children are more likely to be misidentified during biometric
screening.
Of course, we know that other groups, like we have
discussed today, people of color, seniors, are also
misidentified.
Mr. Wagner, what actions is CBP taking to correct the
patterns of errors identified in the NIST report?
Mr. Wagner. Well, again, we are using a high-performing
algorithm that we are not seeing those demographic-based error
rates.
Now, if someone does not match to either the gallery or to
the document they are presenting, we will physically examine
the document.
Ms. Underwood. Right.
Mr. Wagner. Manually look at the picture, and if we have
the confidence it is the person, we can do that through
questioning. We could do that through additional forms of
identification. We can do that through an inspection of the
person.
Sometimes it is just looking at the passport and going,
``OK. That is you. Go ahead.''
It all depends on how discrepant you look from your travel
document photograph.
Ms. Underwood. Some passengers report being unaware or
confused about how to opt out of their biometric screening. As
CBP expands the biometric screening program, does it intend to
reevaluate the best method of communicating the important opt
out information to passengers?
Mr. Wagner. Yes. So right now we have got signage at the
airports, but you know, a lot of people do not read signs at
the airport.
We have got gate announcements that the airlines try to
make before boarding, but again, there is always competing
announcements going on, and sometimes it is tough to understand
what is being said.
So we are actually looking with the airlines as could we
print things on the boarding pass.
Could we give notifications when they are, say, booking
their ticket or when they are getting their check-in
information for boarding?
Are there electronic messages we could provide?
Ms. Underwood. Right.
Mr. Wagner. So we are looking at additional ways to do
that.
We also started taking out some privacy advertisements
advising people of the requirements and what their options are
as well, too.
Ms. Underwood. OK. Well, it is certainly my interest in
making sure that every passenger understands that, No. 1, this
is happening and, No. 2, that they have a choice to opt out,
and I would certainly urge the CBP to strongly consider and
issue this committee a time line for perhaps outlining how we
can improve that communication to all passengers.
Does CBP capture and report the rate of false positives or
mistaken identifications among different demographics at each
port of entry where biometric technology is used?
Mr. Wagner. What we track are the people that we take a
photograph of or receive a photograph of.
Ms. Underwood. Right.
Mr. Wagner. And we are not able to match it to their travel
document that is in our gallery.
Ms. Underwood. Right.
Mr. Wagner. Again, that is that 2 to 3 percent.
Our review of that information does not show noticeable
discrepancies on any types of----
Ms. Underwood. That was not my question. My question is
capturing and reporting by port of entry. So we want to know
the false positives. Are we seeing more at certain places along
the border?
Are we seeing more false positives at certain airports?
Mr. Wagner. We are not seeing false positives that is
matching you to a different identity. We are not seeing that
with this technology.
Ms. Underwood. Or mistaken identities?
Mr. Wagner. We are not seeing that. We are more likely you
do not match against anything. So we get a no information
return.
Ms. Underwood. OK. Dr. Romine, can you elaborate on what
NIST recommends to algorithm developers to improve accuracy
across demographics?
Mr. Romine. The report, the testing that we do does not
result in recommendations specifically to the vendors other
than to take the data that we provide, the evaluation results,
and strive to use those results to improve their methods. But--
--
Ms. Underwood. So you are saying that you do not have a lot
of interaction with the developers?
Mr. Romine. We have informal interaction with them in the
sense that the scientists who do this biometric testing are
part of a larger biometrics community. We see the vendor
representatives, the scientists at meetings, and so on.
But with regard to the FRVT itself, the testing, the
feedback that we provide to the vendors is the test result.
Ms. Underwood. OK. So you all are not doing like convenings
with industry and helping them improve the quality of their
product?
Mr. Romine. We do host events, but more as a convener to
get the community together to discuss different techniques. But
we do not provide, other than sort-of in the general scientific
community sense, we do not provide specific recommendations for
their improvement.
Ms. Underwood. OK. I recognize my time has expired. We
would just like to get more information about that in writing.
Mr. Romine. Happy to do that.
Ms. Underwood. Thank you, sir.
Chairman Thompson. Thank you very much.
The Chair recognizes the gentlelady from New Jersey, Mrs.
Watson Coleman.
Mrs. Watson Coleman. Thank you, Mr. Chairman.
Thank you for your testimony.
A couple of questions. I think I want to just talk more
about the role of the CRCL and NIST. It seems to me that there
has not been much coordination across the DHS spectrum of
directions from DHS to each component regarding their
deployment of biometric technologies. You can correct me if I
am wrong.
Is there any sort of Department-wide strategy in place for
the use of biometric technologies or are components like yours
given wide latitude to stand up biometric programs as you
please?
Mr. Wagner. I am sorry?
Mrs. Watson Coleman. Are you a Lone Ranger?
Mr. Wagner. Are we what? I am sorry. I did not hear that.
Mrs. Watson Coleman. OK. It does not seem like there is
coordination. It does not seem like there is this sort-of
Department-wide oversight.
I want to know whether or not you are getting directions
from others because then I am going to ask Mr. Mina what is
your role and to what degree have you been involved in the
oversight and in signing off on how these things are being
done.
Mr. Wagner. Yes. So as we build out new programs, we are
bound by certain statutes that require us to publish, say, your
Systems of Record notice, your privacy impact assessment,
where, you know, things are reviewed by, you know, our internal
counsel or our Privacy Officer, and to make sure we make and
meet all of the requirements of the statutes.
Do you have the authority to collect what you are doing?
You know, is your time line for storing it and sharing it,
is that all permissible in law?
Is it consistent with your mission? Are you authorized to
do those things?
Mrs. Watson Coleman. So are you operating within your sort-
of silo?
This is what the law says with regard to what you can do.
Is this how you execute based upon what your interpretation is
of that or is there a DHS component that plays into this as
well and says, ``OK. But this is how we want to see this''?
Mr. Wagner. Well, depending on like the acquisition
process, there is a multitude of people at DHS that look at the
acquisition, the resources spent.
There is a whole process to go through for approval before
various boards that authorize the expenditures and the
investment in that. There is the DHS privacy officer. There is
DHS counsel. So there is a lot of oversight by DHS already in
this process.
Certainly the rulemakings would go through with DHS
counsel, with DHS policy. So there is a lot of oversight and
coordination.
Mrs. Watson Coleman. It is my understanding though that
there is no centralized body within the Department that gives
the program a stamp of approval or certifies that they are
ready for prime time. Is that correct?
Has CRCL approved your program? Do you know?
Mr. Wagner. No. They would not necessarily go to them for
approval, but there is----
Mrs. Watson Coleman. Well, for approval in the sense of
maintaining or protecting privacy rights.
Mr. Wagner. So things are reviewed by them, and I will
refer to my colleague.
Mrs. Watson Coleman. What authority do you have, sir, Mr.
Mina?
Mr. Mina. Why do I not answer that in a couple of different
ways, Congresswoman?
So let me step back a second and talk a little bit about
the first part of your question regarding sort-of the
enterprise-level review.
I think one of the ways in which CRCL participates in that
dialog is by serving on enterprise level-wide working groups
across the Department that include representatives from CDP,
DHS S&T, and the Office of Biometric Identity Management, where
we actually are talking about a lot of these issues.
Now, we do not have a privacy impact assessment type model.
However, we do work very closely with the Privacy Office
regarding not just facial recognition technology but certainly
other forms of biometric identification that the Department
uses.
Now, with regard to our relationship with CBP, we work with
them in a couple of different ways. First is very, you know,
directly in terms of offering them advice and then also we on-
site visits and we also work with CBP and the Privacy and Civil
Liberties Oversight Board and their engagement as well.
Mrs. Watson Coleman. So is your role anything more than
just advice, observation and advice?
You have no authority to say, ``No, that is not working.
That is a violation.'' No, that is it, right? Advice?
Mr. Mina. That is not entirely accurate. What I would say
is, yes, we do have an advisory capacity. We also have a
compliance function where we do offer recommendations to
components based on the----
Mrs. Watson Coleman. If they do not follow them?
Mr. Mina. Then we can elevate it if necessary.
Mrs. Watson Coleman. OK. I have one last question?
Chairman Thompson. Yes.
Mrs. Watson Coleman. The question has to do with just the
whole system that is used when we are taking pictures and you
know.
Who is in charge of determining whether or not the lighting
is good, the background is adequate, the cameras are good, they
are placed right so that we can get the best pictures that we
need to get?
Is there anyone in charge of that?
Mr. Wagner. CBP would be, and that is going to be based on,
you know, our results of, say, the match rates. You know, you
can have an airport with a bank of booths and the windows are
such that the sunlight comes in and affects these booths during
the morning and these booths in the afternoon. Those are the
things we have got to look at as we deploy this.
What are the environmental factors that are going to
influence all the different locations that we are going to do
this?
Then we try to adjust, and that might mean we add more tint
to the windows.
Mrs. Watson Coleman. How do you do that? How do you do
that?
Mr. Wagner. We do that internally by reviewing the data and
the results of what happens.
Mrs. Watson Coleman. On what kind of a basis? Weekly?
Daily? Monthly? Whatever.
Mr. Wagner. All of it.
Mrs. Watson Coleman. You know what time the sun comes in
that window, and you know what time the sun comes in that
window.
Mr. Wagner. Right. I would say we do it continuously to get
the best production we can out of it.
Mrs. Watson Coleman. Thank you. Thank you very much.
Thank you, Mr. Chairman.
Chairman Thompson. Thank you very much.
The Chair recognizes the Ranking Member.
Mr. Rogers. Thank you, Mr. Chairman.
I have a unanimous consent request that two articles on
this topic be admitted to the record.
Chairman Thompson. Without objection, so ordered.
[The information follows:]
White Paper by Security Industry Association
what nist data shows about facial recognition and demographics
By: Jake Parker, Senior Director of Government Relations, Security
Industry Association, [email protected]
introduction
In December 2019, the National Institute of Standards and
Technology (NIST) published the most comprehensive report\1\ to date on
the performance of facial recognition algorithms--the core component of
facial recognition technology--across race, gender, and other
demographic groups. The most significant takeaway from the NIST report
is that it confirms current facial recognition technology performs far
more effectively across racial and other demographic groups than had
been widely reported; however, we've seen some misleading conclusions
drawn from the highly technical 1,500-page report. A closer look at the
findings in their proper context is essential to understanding the
implications.
---------------------------------------------------------------------------
\1\ Patrick Grether, Mei Ngan, and Kayee Hanaoka, Face Recognition
Vendor Test (FRVT) Part 3: Demographic Effects (Washington, DC:
National Institute of Standards and Technology, December 2019), https:/
/nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8280.pdf#page=69.
---------------------------------------------------------------------------
key takeaways
Facial recognition technology performs far more effectively
across racial and other demographic groups than widely
reported.
The most accurate technologies displayed ``undetectable''
differences between demographic groups, calling into question
claims of inherent bias.
Key U.S. Government programs are using the most accurate
technologies.
Accuracy rates should always be considered in application-
specific contexts.
role of nist in facial recognition evaluation
For the past 20 years, NIST's Face Recognition Vendor Test (FRVT)
program has been the world's most respected evaluator of facial
recognition algorithms--examining technologies voluntarily provided by
developers for independent testing. NIST's December report is the most
comprehensive scientific evaluation to date of client facial
recognition technology performance across demographic variables,
involving 189 algorithms from 99 developers using 18 million images of
8 million people within 4 different data sets. The results are a
snapshot in time, providing a critical benchmark against which
developers work to improve the technology, as industry progress is
tracked through the on-going FRVT program.
Purpose of the Report and What it Found
NIST's report addresses ``assertions that demographic dependencies
could lead to accuracy variations and potential bias''\2\ as well as
flaws in prior research and media reporting. ``Much of the discussion
of face recognition bias in recent years cites two studies showing poor
accuracy of face gender classification algorithms on black women. Those
studies did not evaluate face recognition algorithms, yet the results
have been widely cited to indict their accuracy,'' according to the
report.\3\ The most-cited figure from those papers is that 2 such
algorithms assigned the wrong gender to photos from that demographic
group nearly 35 percent of the time. This was reported widely in media
reports as a groundbreaking discovery on facial recognition accuracy
even though it did not even assess this technology.
---------------------------------------------------------------------------
\2\ See Demographic Effects, pg. 1.
\3\ See Demographic Effects, pg. 4.
---------------------------------------------------------------------------
In contrast, NIST found that, ``To the extent there are demographic
differentials, they are much smaller,'' pointing out error rates in
verification-type algorithms are ``absolutely low,'' generally below 1
percent and many below 0.5 percent.\4\ Even more significantly, NIST
found that in the most accurate algorithms it tested, differences in
performance across demographic groups were ``undetectable.'' It would
not be possible to mitigate these effects if bias is inherent in facial
recognition technology, as some have alleged.
---------------------------------------------------------------------------
\4\ See Demographic Effects, pg. 54.
---------------------------------------------------------------------------
Notably for policy makers, the most well-known U.S. Government
applications already use some of the highest-performing technologies.
The report specifically identifies 6 suppliers of identification-type
algorithms with undetectable differences in ``false positive''
rates.\5\ Included among these are current technology suppliers to the
Federal Bureau of Investigation Criminal Justice Information Services
Division and U.S. Customs and Border Protection's Traveler Verification
Service.
---------------------------------------------------------------------------
\5\ See Demographic Effects, pg. 8.
---------------------------------------------------------------------------
For the rest of the algorithms, the report found that higher
overall accuracy means smaller differences in performance across
demographic groups. NIST did find relatively higher false positive
effects for some groups in the majority of algorithms tested--depending
on the specific metric, type of algorithm, chosen similarity score
threshold and data set involved. However, as one recent analysis of the
report noted ``Algorithms can have different error rates for different
demographics but still be highly accurate.''\6\
---------------------------------------------------------------------------
\6\ Michael McLaughlin and Daniel Castro, ``The Critics Were Wrong:
NIST Data Shows the Best Facial Recognition Algorithms are Neither
Racist Nor Sexist,'' Information Technology and Innovation Foundation,
Jan. 27, 2020, pg. 3, https://itif.org/publications/2020/01/27/critics-
were-wrong-nist-data-shows-best-facial-recognition-algorithms.
---------------------------------------------------------------------------
NIST charts comparisons across demographic groupings on a
logarithmic scale because this granularity allows us to better perceive
relative differences between error rates produced by algorithms that
may be highly accurate in absolute terms. According to NIST, ``readers
don't perceive differences in numbers near 100 percent well,'' due to
the ``high nineties effect where numbers close to 100 are perceived
indifferently.''\7\
---------------------------------------------------------------------------
\7\ See Demographic Effects, pg. 22.
---------------------------------------------------------------------------
As a result, some figures in the report appear large if considered
only in relative terms. Using photos from over 24 countries in 7
distinct global regions, verification-type algorithms produced false
match rates for photos of individuals originally from East Africa as
much as ``100 times greater than baseline.'' Although performance
variations across demographic groups are important to continually
assess and critically examine, outside of Somalia nearly all country-
to-country comparisons across algorithms yielded false match rates of
less than 1 percent \8\ despite the magnitude of differences
identified.
---------------------------------------------------------------------------
\8\ See Demographic Effects, Annex 7.
---------------------------------------------------------------------------
Similarly, only 4 out of 116 algoritluns tested using the U.S.
Mugshot Identification Database had false match rates of more than 1
percent for any demographic: Male, female, black, white, Asian, or
American Indian.\9\ One example cited by NIST produced a 0.025 percent
false match rate for black males and a 0.1 percent false match rate for
black women.\10\ Compared to the rate for white males, this is 10 times
higher for black women and 2.5 times higher for black males; however,
these error rates are at or below \1/10\ of 1 percent.
---------------------------------------------------------------------------
\9\ See Demographic Effects, Annex 6.
\10\ See Demographic Effects, pg. 46, figure 12, imperial--002.
---------------------------------------------------------------------------
Certainly, significant gaps were found between the very highest-
and lowest-performing algorithms. NIST tests any algorithm submitted
and many of these are in the early stages of development. Lower-
performing technologies are less likely to be deployed in commercial
products.
accuracy in context
There will always be error rates for any biometric, or any
technology for that matter. For example, this is why NIST compared
false match rates for different demographic groups to each other, not
zero. How is accuracy defined when it comes to demographic effects?
According to NIST, it means these rates ``do not vary (much) over any
demographics.''\11\
---------------------------------------------------------------------------
\11\ See Demographic Effects, pg. 74.
---------------------------------------------------------------------------
Overall, modern facial recognition technology is highly accurate.
It is in fact image quality variations like pose, illumination, and
expression have been the primary driver of errors in facial recognition
performance, not demographic effects, and growing immunity to such
problems is, according to NIST, the ``fundamental reason why accuracy
has improved since 2013.''\12\
---------------------------------------------------------------------------
\12\ Patrick Grother, Mei Ngan and Kayee Hanaoka, Face Recognition
Vendor Test (FRVT) Part 2: Identification (Washington, DC: National
Institute of Standards and Technology, September 2019), pg. 8, https://
www.nist.gov/system/files/documents/2019/09/11/nistir_8271_-
20190911.pdf.
---------------------------------------------------------------------------
NIST has documented massive improvements in recent years, noting in
2018 \13\ the software tested was at least 20 times more accurate than
it was in 2014, and in 2019 \14\ finding ``close to perfect''
performance by high-performing algorithms with miss rates averaging 0.1
percent. On this measurement, the accuracy of facial recognition is
reaching that of automated fingerprint comparison, which is generally
viewed as the gold standard for identification.\15\
---------------------------------------------------------------------------
\13\ NIST Evaluation Shows Advance in Face Recognition Software's
Capabilities, (Washington, DC: National Institute of Standards and
Technology, November 2018), https://www.nist.gov/news-events/news/2018/
11/nist-evaluation-shows-advance-face-recognition-softwares-
capabilities.
\14\ Patrick Grother, Mei Ngan, and Kayee Hanaoka, Face Recognition
Vendor Test (FRVT) Part 2: Identification (Washington, DC: National
Institute of Standards and Technology, September 2019), pg. 6, https://
www.nist.gov/system/files/documents/2019/09/11/nistir_8271-
_20190911.pdf.
\15\ See NIST's most recent fingerprint vendor technology
evaluation of the most accurate submissions for ten finger (rolled-to-
rolled) samples, https://nvlpubs.nist.gov/nistpubs/ir/2014/
NIST.IR.8034.pdf.
---------------------------------------------------------------------------
lab tests vs. real-world
We simply aren't seeing instances in the United States where
demographic performance differences in widely-used algorithms are
affecting facial recognition systems in high-risk settings. There are
several reasons that may explain why.
Algorithms comprise just one of several components of facial
recognition systems. A human analyst will play a critical role in use
of facial recognition as a tool in law enforcement investigations or as
part of any process with potential high-consequence outcomes for
individuals. There are no automated decisions made solely by the
technology in these cases. Personnel adjudicates in situations where
the technology may not work as well as intended. NIST has documented
that the most accurate identification results occur when facial
recognition is combined with trained human review, versus either
element alone.\16\ This may explain U.S. law enforcement's decade-plus
operating history without any example of it contributing to a mistaken
arrest or imprisonment.
---------------------------------------------------------------------------
\16\ NIST Study Shows Face Recognition Experts Perform Better With
AI as Partner, (Washington, DC: National Institute of Standards and
Technology, May 2018), https://www.nist.gov/news-events/news/2018/05/
nist-study-shows-face-recognition-experts-perform-better-ai-partner.
---------------------------------------------------------------------------
False positives are naturally limited by the size of the data set
used. A larger set of photos likely has a larger number of similar
people in it; however, for many applications, the data sets are
relatively small--the 250 passengers on a flight or 2 dozen people
authorized to enter a building, for example, which will naturally limit
false positives.
NIST calls for considering different accuracy measurements within
the context of the ``performance metric of interest'' for specific
applications, noting the study is the first to ``properly report and
distinguish between false positive and false negative effects.''\17\
The real-world implications of each depend entirely upon the specific
use and mitigating factors. An error could be mostly inconsequential in
cases where a ``subject experiencing a false rejection could make a
second attempt at recognition''\18\ in order to unlock a door or device
or clear passport control, for example.
---------------------------------------------------------------------------
\17\ See Demographic Effects. pg. 18.
\18\ See Demographic Effects, pg. 58.
---------------------------------------------------------------------------
One of the report's key findings was that false positive rates vary
much more across demographic groups than false negative effects;
however, false negative effects are more critical to many uses
identified.\19\ For example, facial recognition is used to detect fraud
attempts when the same person applies for driver's license applications
under different identities, ensuring this person is not the same as any
other in a database. This is also how it works in many security
applications, where the purpose of photo comparison is to ensure
persons entering a building do not match those on a persons of interest
list. In both cases, the false negative rate is the key performance
measurement because the antifraud or security objective requires a very
low likelihood of missing a possible match to flag for human review.
---------------------------------------------------------------------------
\19\ See Demographic Effects, charts on pgs. 29, 62.
---------------------------------------------------------------------------
For law enforcement investigations, ensuring that possible matches
are not missed is even more critical. According to the NIST report,
``false positive differentials from the algorithm are immaterial'' for
law enforcement investigations since all searches produce a fixed
number of candidates for human review regardless of any threshold for
similarity score.\20\ On the other hand, at a port of entry, there may
be a relatively high risk of persons attempting to enter under another
identity, so false positive effects may be more critical. In a low-risk
application like entry to an amusement park, both accuracy measurements
may be less critical due to the low probability of someone trying to
impersonate someone with a ticket and the operational need to speed
entry by limiting rejections.
---------------------------------------------------------------------------
\20\ See Demographic Effects, pg. 5.
---------------------------------------------------------------------------
limitations of the report
Despite taking the most comprehensive look so far at demographic
effects in facial recognition performance, the NIST report does have
limitations and raises some unanswered questions. Most significantly,
it is not clear whether ethnicity was fully isolated from other
demographics or capture conditions in many instances. For example,
false match rates for Somalia are very significant outliers that are
not fully explained. These error rates are far higher for Somalians
than neighboring countries in nearly every algorithm tested. For
example, one of the most accurate verification algorithms overall had a
false match rate of about 1 percent for Somalia, while for neighboring
Ethiopia--which has a closely related ethnic majority--it was just 0.07
percent, more than 14 times lower.\21\ This dramatic difference would
suggest that the impact of ethnicity was not isolated and that other
differences, in capture conditions, data labeling errors, etc. between
country data exist.
---------------------------------------------------------------------------
\21\ See Demographic Effects, Annex 7, pg. 226, tevian-005.
---------------------------------------------------------------------------
implications for the security industry
Applied to security solutions developed by our industry, biometric
technologies like facial recognition increase the effectiveness of
safety and security measures that protect people from harm. Any
significant bias in technology performance makes it harder to achieve
this goal.
We understand that there are legitimate concerns that use of facial
recognition technology might negatively impact women and minorities.
Industry is striving to provide technology that is as effective and
accurate as possible across all types of uses, deployment settings and
demographic characteristics in order to fully address these concerns.
Both developers and end-users have a responsibility to minimize any
negative effects that could result when the technology does not perform
as intended though proper design, configuration, policies, and
procedures. We strongly believe that facial recognition makes our
country safer and brings value to our everyday lives when used
effectively and responsibly. No technology product should ever be used
for purposes that are unlawful, unethical, or discriminatory.
FACE FACTS: HOW FACIAL RECOGNITION MAKES US SAFER & THE DANGERS OF A
BLANKET BAN
Facial recognition technology makes our country safer and brings
value to our everyday lives when used effectively and responsibly. The
Security Industry Association (SIA) believes all technology products,
including facial recognition technology, must only be used for purposes
that are lawful, ethical, and nondiscriminatory.
Modern facial recognition technology is highly accurate. The
National Institute of Standards and Technology (NIST) found
that the facial recognition software it tests is now over 20
times better than it was in 2014 at searching a database to
find a matching photograph. NIST's September 2019 report found
``close to perfect'' performance by high-performing algorithms
with miss rates averaging 0.1 percent, reaching the accuracy of
fingerprint comparison technology--the gold standard for
identification.
The benefits of facial recognition have been proven for more
than a decade of use in real-world applications, including
finding missing and exploited children, protecting critical
infrastructure, and aiding law enforcement investigations. See
examples of the benefits in action on the reverse page.
why a blanket ban puts americans at risk
A blanket ban on Government use precludes all possible
current and future applications of the technology, regardless
of the purpose, putting the safety of every resident at risk.
Beyond law enforcement, such a ban prohibits other proven
uses like secured employee access to critical infrastructure
and other systems that protect building occupants and software
that detects fraud against Government programs, to name a few.
Such bans have also been defined broadly, prohibiting any
Government official, employee, contractor or vendor from using
any technology with facial recognition capabilities, including
social media platforms and smartphones.
A ban on facial recognition eliminates a useful tool that is
being used alongside human intelligence. Thorough analysis must
acknowledge the alternatives a ban would leave us with--far
slower and less accurate identification processes chat are much
more prone to errors (for example, detectives sifting manually
through hundreds or even thousands of videos and images of
arrested individuals based on suspect descriptions). NIST
confirmed in a 2018 study chat the highest identification
accuracy is achieved through human analysis supported by facial
recognition technology versus either element alone.
Before taking such an extreme step, policy makers must
thoroughly examine how the technology is used and consider all
the options available to address concerns. Sensible
transparency and accountability measures can be identified that
would ensure responsible use of the technology without
unreasonably restricting cools chat have become so essential to
public safety.
FACE FACTS: KEEPING AMERICANS SAFE
SAVING SEX TRAFFICKING VICTIMS.--In April 2019, a California law
enforcement officer saw a social media post about a missing child from
the National Center for Missing and Exploited Children. The officer
used facial recognition which returned a list of on-line sex ads
featuring the girl.
According to a story in WIRED, the girl had been ``sold for
weeks,'' and the officer's actions helped a process that ``recovered
and removed from the girl from trauma.''
CATCHING A NEW YORK CITY SUBWAY TERRORIST.--In August 2019, New
York Police Department detectives used facial recognition to help
identify a man who sparked terror by leaving rice cookers in and around
a subway station. Detectives pulled still images from security footage
and used facial recognition software, along with additional
investigative work, to identify the suspect within an hour: NYPD
officials were quoted saying, ``To not use technology like this would
be negligent'' and ``This is the most important type of case that we'd
see out there: a possible terrorist attack in NYC.''
FINDING A KILLER WHO TARGETED LGBTQ VICTIMS.--On May 25, 2019, in
Wayne County, Michigan, 3 members of the LGBTQ community were shot and
killed by a man at a gas station. The Detroit Police Department used
facial recognition, as well as their own intelligence, to help identify
the suspect, who was charged with 3 counts of murder in addition to
other charges.
IDENTIFYING THE CAPITAL GAZETTE KILLER.--Jarrod Ramos was angered
by a story the Capital Gazette Newspaper in Annapolis, Maryland, ran
about him in 2011 and brought a lawsuit against the paper for
defamation, which a judge later dismissed. In June 2018, Ramos entered
the newspaper building with a shotgun and killed 5 employees, leaving 2
others critically injured. Anne Arundel Police obtained an image of
Ramos and sent it to the Maryland Combined Analysis Center, which
helped identify him by comparing the photo to others in the Maryland
Image Repository System.
APPREHENDING PEDOPHILES EVADING JUSTICE.--In 2017, after a 16-year
manhunt, a man accused of sexually assaulting a minor was apprehended
in Oregon. Using facial recognition technology, the Federal Bureau of
Investigation (FBI) was able to identify the suspect after a positive
match was found when the suspect sought to acquire a U.S. passport.
Similarly, in 2014, the FBI used facial recognition technology to help
locate and apprehend a convicted pedophile who had been on the run for
14 years.
PREVENTING ENTRY INTO THE UNITED STATES UNDER FALSE IDENTITIES.--
After just 3 days of operation, facial recognition technology at Dulles
International Airport in Virginia caught a man trying to use a fake
passport to enter the United States. The fraudulent passport would have
easily gone undetected with visual inspection alone. The ability to
enter under a false identity is essential to organized crime, human
trafficking, money laundering, drug smuggling, terrorism, and many
other criminal activities. According to U.S. Customs & Border
Protection, use of the technology prevented 26 alleged imposters from
entering the United States in just a 3-month span in 2018.
______
Article From the Information Technology and Innovation Foundation
(ITIF)
The Critics Were Wrong: NIST Data Shows the Best Facial Recognition
Algorithms Are Neither Racist Nor Sexist
By: Michael McLaughlin and Daniel Castro/January 2020
A close look at data from a new NIST report reveals that the best
facial recognition algorithms in the world are highly accurate and have
vanishingly small differences in their rates of false positive or
false-negative readings across demographic groups.
introduction
The National Institute of Standards and Technology (NIST) recently
released a report that examined the accuracy of facial recognition
algorithms across different demographic groups. The NIST report found
that the most accurate algorithms were highly accurate across all
demographic groups. But NIST tested nearly 200 algorithms from vendors
and labs around the world--it allows anyone to submit an algorithm for
testing--and since many of the algorithms it tested displayed some
bias, several news outlets and activists have misleadingly concluded
that facial recognition systems are racist and sexist.\1\ But a close
look at the data reveals a different picture.
---------------------------------------------------------------------------
\1\ Tom Higgins `` `Racist and Sexist' Facial Recognition Cameras
Could Lead to False Arrests,'' The Telegraph, December 20, 2019,
https://www.telegraph.co.uk/technology/2019/12/20/racist-sexist-facial-
recognition-cameras-could-lead-false-arrests.
---------------------------------------------------------------------------
Facial recognition technology compares images of faces to determine
their similarity, which the technology represents using a similarity
score. The technology often performs one of two types of comparisons.
The first comparison is known as a one-to-many or identification
search, in which the technology uses a probe image to search a database
of images to find potential matches. The second comparison is known as
a one-to-one or verification search as the technology compares 2 images
to determine the similarity of the faces in them. In many cases, the
faces in images are considered a match if their similarity score meets
or exceeds the match threshold, a number the operator assigns that
represents a minimum acceptable similarity score. The technology has
many commercial and non-commercial uses, and will likely be integrated
into more products and services in the future to enhance security,
improve convenience, and increase efficiency. such as by helping find
victims of human trafficking. expediting passengers through airport
security, and flagging individuals using forged identification.\2\
---------------------------------------------------------------------------
\2\ Tom Simonite, ``How Facial Recognition Is Fighting Child Sex
Trafficking,'' Wired, June 19 2019, https://www.wired.com/story/how-
facial-recognition-fighting-child-sex-trafficking/; ``Face Recognition
Nabs Fake Passport User at US Airport,'' VOA News, August 24, 2018
https://www.voanews.com/silicon-valley-technology/face-recognition-
nabs-fake-passport-user-us-airport.
---------------------------------------------------------------------------
NIST assessed the false positive and false-negative rates of
algorithms using 4 types of images, including mugshots, application
photographs from individuals applying for immigration benefits, visa
photographs, and images taken of travelers entering the United States.
NIST's report reveals that:
The most accurate identification algorithms have ``
undetectable'' differences between demographic groups;\3\
---------------------------------------------------------------------------
\3\ We defined the most accurate identification algorithms as the
20 algorithms that had the lowest false-negative identification rates
for placing the correct individual at rank one when searching a
database that had images of 12 million individuals in NIST's September
2019 identification report. NIST provided error characteristics data by
race and sex for 10 of these algorithms in its recent report.
Consequently, we analyzed the performance of NEC-2, NEC-3, Visionlabs-7
, Microsoft-5, Yitu-5, Microsoft-0, Cogent-3, ISystems-3,
NeuroTechnology-5, and NTechlab-6; Patrick Grother, Mei Ngan, and Kayee
Hanaoka, Face Recognition Vendor Test (FRVT) Part 2: Identification
(Washington, DC: National Institute of Standards and Technology,
September 2019), 47 https://www.nist.gov/system/files/documents/2019/
09/11/nistir_8271_20190911.pdf#page=49.
---------------------------------------------------------------------------
The most accurate verification algorithms have low false
positives and false negatives across most demographic
groups;\4\
---------------------------------------------------------------------------
\4\ We defined the most accurate verification algorithms as those
that rank in the top 20 on NIST's FRVT 1:1 leaderboard on January 6,
2020. NIST has since updated the leaderboard. Not all of these
algorithms were tested in NIST's most recent demographics report. We
analyzed the performance of algorithms that NIST provided data for in
Annexes 6, 13, 15, and Figure 22. These algorithms are visionlabs-007,
everai-paravision-003, didiglobalface-001, imperial-002, dahua-003,
tevian-005, alphaface-001, ntechlab-007, yitu-003, innovatrics-006,
facesoft-000, intellifusion-001, anke-004, hik-001, camvi-004, vocord-
007, and tech5-003; National Institute of Standards and Technology,
FRVT 1:1 Verification (FRVT 1: 1 leaderboard, accessed January 6,
2020), https://pages.nist.gov/frvt/html/frvt11.html.
---------------------------------------------------------------------------
Algorithms can have different error rates for different
demographics but still be highly accurate.
key findings
As detailed below, NIST found that the most accurate algorithms--
which should be the only algorithms used in Government systems--did not
display a significant demographic bias. For example, 17 of the highest-
performing verification algorithms had similar levels of accuracy for
black females and white males: False-negative rates of 0.49 percent or
less for black females (equivalent to an error rate of less than 1 in
200) and 0.85 percent or less for white males (equivalent to an error
rate of less than 1.7 in 200).\5\
---------------------------------------------------------------------------
\5\ The high-performing algorithms include visionlabs-007, everai-
paravision-003, didiglobalface-001, imperial-002, dahua-003, tevian-
005, alphaface-001, ntechlab-007, yitu-003, innovatrics-006, facesoft-
000, intellifusion-001, anke-004, hik-001, camvi-004, vocord-007, and
tech5-003. Comparisons made at the same match threshold.
---------------------------------------------------------------------------
While the most accurate algorithms did not display a significant
demographic bias, it is also true that the majority of the algorithms
NIST tested generally performed better on men and individuals with
lighter skin tones. However, it is important to recognize that there is
a stark difference between the best and worst algorithms. In comparison
to the false-negative rates under 1 percent for black females and white
males among the highest-performing algorithms, the lowest-performing
algorithms had false-negatives rates, for blacks and whites, as high as
99 percent.\6\ This wide range of accuracy is not surprising
considering that NIST allows anyone to submit an algorithm for testing,
ranging from large companies with production systems to small research
groups whose algorithms have not left the lab--algorithms are tested
even if they are not incorporated into a commercially-available
product.
---------------------------------------------------------------------------
\6\ The low-performing algorithms, according to their performance
for false non-match rate on Figure 22 of the NIST demographics report,
include shaman-001, isap-001, ayonix-000, amplifiedgroup-001, saffe-
001, videonetics-001, and chtface-001.
---------------------------------------------------------------------------
The Most Accurate Identification Algorithms Have Undetectable
Differences Between Demographics
NIST found that some highly-accurate algorithms had false-positive
demographic differentials that were so small as to be ``undetectable''
for one-to-many searches.\7\ Moreover, for most algorithms, black men
had lower false-negative rates than white men, and several of the top
algorithms had better false-negative rates for white women than white
men.\8\ Several algorithms also provided uniform similarity scores
across demographic groups, meaning that the algorithms provided similar
match and non-match scores regardless of race and gender.\9\ The
uniform scores indicate that these algorithms would have small
demographic differentials if an operator applied a threshold. But
different thresholds can affect demographic differentials. For example,
at least 6 of the most accurate identification algorithms had higher
false-positive rates for black men than white men at one threshold, but
lower false-positive rates for black men than white men at another
threshold.\10\
---------------------------------------------------------------------------
\7\ Patrick Grether, Mei Ngan, and Kayee Hanaoka, Face Recognition
Vendor Test (FRVT) Part 3: Demographic Effects (Washington, DC:
National Institute of Standards and Technology, December 2019), 3,
https://nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8280.pdf#page=6.
\8\ To compare white men and white women, we analyzed false-
negative rates for ranking the correct matching image as the top
potential match. Algorithms that had lower false-negative rates for
white women than white men include NEC-2, NEC-3, and Visionlabs-7;
Patrick Grether, Mei Ngan, and Kayee Hanaoka, Face Recognition Vendor
Test(FRVT) Part 3: Demographic Effects (Washington, DC: National
Institute of Standards and Technology, December 2019), 63 https://
nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8280.pdf#page=66; National
Institute of Standards and Technology, On-going Face Recognition Vendor
Test (FRVT) (part 3: demographic effects, annex 16: identification
error characteristics by race and sex), https://pages.nist.gov/frvt/
reports/demographics/annexes/annex_16.pdf.
\9\ Patrick Grether, Mei Ngan, and Kayee Hanaoka, Face Recognition
Vendor Test (FRVT) Part 3: Demographic Effects (Washington, DC:
National Institute of Standards and Technology, December 2019), 66,
https://nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8280.pdf#page=69.
\10\ These algorithms include Visionlabs-7, Microsoft-5, Yitu-5,
Microsoft-0, ISystems-3, and NeuroTechnology-5; National Institute of
Standards and Technology, On-going Face Recognition Vendor Test (FRVT)
(part 3: demographic effects, annex 16: identification error
characteristics by race and sex), https://pages.nist.gov/frvt/reports/
demographics/annexes/annex_16.pdf.
---------------------------------------------------------------------------
The Most Accurate Verification Algorithms Have Low False Positives and
Negatives Across Most Demographics
The most accurate verification algorithms have low false positives
and negatives across most demographics. For example, when NIST applied
thresholds so that the algorithms had false positive rates of 0.01
percent for white males, more than half of the 17 most accurate
algorithms had false-positive rates of 0.03 percent or better for black
males, Asian men, and white women.\11\ This equates to the algorithms
falsely matching these individuals 3 times or less per every 10,000
comparisons to an imposter compared to 1 per every 10,000 for white
males. At another threshold, 7 of the top algorithms displayed no
false-positive bias between white men, black men, Asian men, and white
females.\12\ At this threshold, several algorithms also had false-
positive rates of 0.003 percent or less for black women or Asian women
while white males had false-positive rates of 0.001 percent.\13\
---------------------------------------------------------------------------
\11\ One of these algorithms, for example, is visionlabs-007;
National Institute of Standards and Technology, On-going Face
Recognition Vendor Test (FRVT) (part 3: demographic effects, annex 6:
cross-race and sex false match rates in United States mugshot images),
https://pages.nist.gov/frvt/reports/demographics/annexes/annex_06.pdf.
\12\ An example of such an algorithm is anke-004.
\13\ An example of such an algorithm is yitu-003.
---------------------------------------------------------------------------
False negatives were also low for the most accurate verification
algorithms. Five of the 17 most accurate algorithms had false-negative
rates of less than 1 percent across all demographic groups when NIST
applied a threshold that set false-positive rates at 0.01 percent.\14\
Similarly, the best verification algorithms had less than 1 percent
false-negative rates across countries and demographic groups. For
example, the algorithm Visionlabs-007 had below a 1 percent false-
negative rate for nearly all countries and demographic groups for
border crossing application images. There were two exceptions--Somalian
and Liberian women under 45. Nonetheless, the algorithm had a false-
negative rate below 1.4 percent for each of these groups.
---------------------------------------------------------------------------
\14\ These algorithms are visionlabs-007, everai-paravision-003,
didiglobalface-001, alphaface 001, and intellifusion-001; National
Institute of Standards and Technology, On-going Face Recognition Vendor
Test (FRVT) (part 3: demographic effects, annex 15: genuine and
imposter score distributions for United States mugshots), https://
pages.nist.gov/frvt/reports/demographics/annexes/an nex_15.pdf.
---------------------------------------------------------------------------
Algorithms Can Have Different Error Rates for Different Demographics
But Still Be Highly Accurate
Some algorithms perform differently on one group compared to
another, but still maintain true positive and true negative accuracy
rates greater than 99 percent for all races and sexes.\15\ Because
these algorithms have very low error rates, differences that are small
in absolute terms may seem large if expressed in relative terms. For
example, an algorithm from Dutch firm VisionLabs, Visionlabs-007, had a
false-negative rate 4 times higher for the nationality it performed
poorest on (Somalian) than the nationality it performed best on
(Salvadoran).\16\ Nonetheless, the algorithm only had a false-negative
rate of 0.63 percent for individuals from Somalia. Another example is
the performance difference of a verification algorithm from Camvi, a
firm based in Silicon Valley, for white males and American Indian
females. At one particular threshold, the algorithm had a false-
positive rate that was 13 times higher for American Indian females than
white men.\17\ But at this threshold, the algorithm had barely more
than 1 false match of American Indian females for every 10,000 imposter
comparisons to other American Indian females. It is also true that most
verification algorithms had higher false-negative rates for women than
men. But NIST notes that this ``is a marginal effect--perhaps 98
percent of women are still correctly verified--so the effect is
confined to fewer than 2 percent of comparisons where algorithms fail
to verify.''
---------------------------------------------------------------------------
\15\ These algorithms include visionlabs-007 and everai-paravision-
003.
\16\ In this case, NIST set the threshold to ``the lowest value
that gives FMR less than or equal to 0.00001.''
\17\ National Institute of Standards and Technology, On-going Face
Recognition Vendor Test (FRVT) (part 3: demographic effects, annex 15:
genuine and imposter score distributions for United States mugshots,
19), https://pages.nist.gov/frvt/reports/demographics/annexes/
annex_15.pdf#20.
---------------------------------------------------------------------------
putting nist's data in context
Recent reporting on how law enforcement in San Diego used facial
recognition from 2012-2019 can also help put NIST's data in context. In
2018, various law enforcement entities made 25,102 queries to a
database of 1.8 million mugshot images.\18\ Law enforcement officials
uses of the technology included attempts to determine whether an
individual had a criminal record and attempts to discover the identity
of individuals who lacked identification. These use cases were likely
one-to-many searches. Law enforcement did not track the success of the
program, making it unclear how many false positives or false negatives
the system registered as well as how many mated or non-mated searches--
a search in which an image of the individual was not in San Diego's
database--they performed.
---------------------------------------------------------------------------
\18\ DJ Pangburn, ``San Diego's Massive, 7-Year Experiment With
Facial Recognition Technology Appears to Be a Flop,'' Fast Company,
January 9, 2020, https://www.fastcompany.com/90440198/san-diegos-
massive-7-year-experiment-with-facial-recognition-technology-appears-
to-be-a-flop.
---------------------------------------------------------------------------
But we can consider a few scenarios to make a rough estimate of how
the most accurate algorithms might perform in a city like San Diego,
assuming San Diego's images and hardware were of similar quality to
NIST's.\19\ Under the first scenario, let us assume that all 25,102
probe images law enforcement used had a match in the database of 1.8
million mugshot images (an unlikely event), and that law enforcement
did not apply a threshold to limit false positives or negatives (also
unlikely). NEC-2, the best identification algorithm NIST tested in an
earlier 2019 report, failed to rank the correct candidate as the most
likely match only 0.12 percent of the time when performing a search of
a database containing images of 3 million individuals.\20\ At this
rate, the technology would have succeeded in listing the correct
individual in the San Diego search as the most likely match 24,970
times out of the 25,000 searches and failed 30 times.
---------------------------------------------------------------------------
\19\ In each of the scenarios, we are assuming that the racial and
gender makeup of San Diego's mugshot database is similar to NIST's
mugshot database.
\20\ Patrick Grother, Mei Ngan, and Kayee Hanaoka, Face Recognition
Vendor Test (FRVT) Part 2: Identification (Washington, DC: National
Institute of Standards and Technology, September 2019), 47, https://
www.nist.gov/system/files/documents/2019/09/11/nistir_8271_-
20190911.pdf#page=49.
---------------------------------------------------------------------------
Under a second scenario, let us assume law enforcement applied a
threshold that allowed for 1 false positive every 1,000 non-mate
searches. At this rate, NEC-3 had a false-negative rate of 0.26
percent. We also assume that half of the more than 25,000 probe images
had a match in the database and that half did not have a match. In this
scenario, the algorithm would have registered 13 false positives and 33
false negatives.
conclusion
Developers and users of facial recognition technology, law
enforcement, and lawmakers can take several actions to promote the
development and responsible use of facial recognition technology.
First, developers should continue to improve accuracy rates across
different demographics, including by diversifying their datasets.\21\
Second, the Government should set standards for the accuracy rates of
the systems it deploys. Third, law enforcement should have standards
for the quality of images it uses in a facial recognition search, which
can affect the accuracy of facial recognition algorithms.\22\ Fourth,
the users of facial recognition technology should carefully choose
which match threshold is appropriate for their goal. Last, lawmakers
should consider how law enforcement typically uses the technology and
the different implications of false positive and false negatives when
developing regulations. In most law enforcement scenarios, law
enforcement is using facial recognition technology to return a list of
possible suspects that humans review. And there are different
implications when algorithms incur false positives or false negatives.
In many cases, a subject can make a second attempt at recognition when
a facial recognition system produces a false negative. This implication
differs from the possible effects of false positives, which could allow
an individual access to a facility they should not enter.
---------------------------------------------------------------------------
\21\ Chinese developers often had lower false positives for Chinese
faces, suggesting that increasing the representation of minority faces
in training data may reduce bias.
\22\ For example, although the false-negative rates were frequently
lowest for black individuals in mugshot images, false-negative rates
were relatively high for black individuals in images taken at border
crossings. This difference could result from inadequate exposure in the
latter photos; Patrick Grother, Mei Ngan, and Kayee Hanaoka, Face
Recognition Vendor Test (FRVT) Part 3: Demographic Effects (Washington,
DC: National Institute of Standards and Technology, December 2019), 54,
https://nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8280.pdf#page=57.
---------------------------------------------------------------------------
Finally, while there is no place for racial, gender, or other types
of discrimination in societies, to ban facial recognition unless it
performs exactly the same across every conceivable group is impractical
and would limit the use of a societally valuable technology. Many
critics of facial recognition technology complain that the technology
is not accurate enough, but refuse to give specifics on what they would
consider sufficient--refusing to set a clear goal post for industry--
which suggests they are not serious about wanting to improve the
technology and oppose it for other reasons.
Reasonable people may disagree on when it is appropriate to use
facial recognition, but the facts are clear that the technology can be
highly accurate. As previous NIST reports have shown, many of the
algorithms have accuracy rates that exceed 99 percent, and as the new
report shows, the differences across demographics are minimal for the
best algorithms.\23\
---------------------------------------------------------------------------
\23\ For example, a previous NIST report revealed that the top 20
identification algorithms failed to place the correct individual as the
top potential match when searching a database containing images of 12
million individuals less than 1 percent of the time; Patrick Grother,
Mei Ngan, and Kayee Hanaoka, Face Recognition Vendor Test (FRVT) Part
2:--Identification (Washington, DC: National Institute of Standards and
Technology, September 2019), 47, https://www.nist.gov/system/files/
documents/2019/09/11/nistir_8271_20190911.pdf#page=49.
Mr. Rogers. Thank you, sir.
Chairman Thompson. The Chair recognizes the gentlelady from
Texas, Ms. Jackson Lee.
Ms. Jackson Lee. Let me thank the Chairman very much.
Let me acknowledge all of the Government witnesses and
thank them for their service.
Let me renew the inquiry that will be pursued by Chairwoman
Rice, but I will add to it, and that is, Mr. Wagner, a better
understanding. Maybe you will provide the information to the
committee on the denial of clear global and Trusted Traveler as
relates to States that may not have the laws that you think are
appropriate or, in the instance of New York, closing out access
to the issue of driver's license.
I raise the question because we should look as the Federal
Government at what other identification options may be valid.
I know that we have known each other for a long time, and I
would think that you would be willing to look at that so that
we can find common ground.
Let me pursue this line of reasoning, and please,
witnesses, understand that I am not saying this is what you are
doing. I need to understand your thinking.
So to the Deputy Executive Assistant Commissioner Wagner,
would you accept the fact that bias could be introduced by
technology if the application developer of the program had a
bias into how an application reacts to different types of
people because it is technology?
Mr. Wagner. Yes.
Ms. Jackson Lee. I would also make the point though it is a
little bit humorous, is I am sure the people in Iowa were
trusting of the app and thought they had something going on
there, and we all can see where we are at this point.
Would you accept that, Mr. Mina?
Mr. Mina. Yes, Congresswoman.
Ms. Jackson Lee. Would NIST accept that, Mr. Romine?
Mr. Romine. Yes.
Ms. Jackson Lee. Yes. An algorithm could be--again, know
that this is not pointed toward you--be written to flag all
black males wearing dreadlocks.
Mr. Wagner, this is in terms of how technology can be.
Mr. Wagner. I guess you could.
Ms. Jackson Lee. I understand. You can say on the record to
your knowledge, you are not using that kind of algorithm.
Mr. Wagner. We are not using that. I can----
Ms. Jackson Lee. That would be very good. I am sure
dreadlock wearers would be glad of that.
Mr. Mina.
Mr. Mina. Yes, that is possible, and again, as Mr. Wagner
said, we have not seen that in our review as well.
Ms. Jackson Lee. All right. Mr. Romine.
Mr. Romine. It is certainly possible.
Ms. Jackson Lee. OK. So here we are. Let me to my
colleagues over here that are DNA advocates, as a member of the
Judiciary Committee, which I had to step away from, we are DNA
lovers. I wrote the Violence Against Women Act and put in $291
million for DNA enhancement. So we understand that as the new
added technology.
But as the Department of Homeland Security, we made a
commitment post-9/11 with George Bush going to the Trade and
saying he heard the firefighters, but at the same time he also
heard Muslims who were indicating it is not the blanket world
of people who happen to be Muslim.
So, in particular, Mr. Mina, I want to try to find out what
aggressive role do you play in helping to not have platitudes.
Forgive me. I am not suggesting you do, but to aggressively
ensure that the biases against black women with dreadlocks, men
with dreadlocks, Muslims or Sikhs wearing attire.
I went through that. I have been on this committee since
its beginning. That is not technology, but and then now
sophisticated technology is not undermining the civil liberties
and civil rights of this Nation and those coming in innocently
to the country. You can use the new technology as well.
Then to Mr. Romine, let me find out how are you continuing
to do your assessment of these algorithms to ensure that it
looks like you were not able to get the exact one that Mr.
Wagner's team is using. That concerns me.
I need you to get every accurate piece of information, and
I would like you to say that.
Mr. Mina, what aggressiveness are you doing to protect the
travelers and the American people?
Mr. Mina. I thank you for the question, Congresswoman.
So I think we are doing a lot of different things across
the spectrum and the life cycle of this program and policy, and
again, I want to focus. Our attention is really on the
application, not so much on the algorithm itself, but on how it
is applied by particularly a DHS program, in this case, CBP.
We do that through, on the policy-making side, working
directly with the component, advising on proposed regulations
of implementing policies, as well as offering suggestions as it
relates to applications, for example, folks wearing religious
headwear or folks that have objections to photography based on
religious reasons or the people who are disabled or otherwise
injured and area not able to take pictures.
We also do it through our robust community engagement. We
talk to members of the community across the country, and, Mr.
Chairman, I actually have the information in front of me
regarding some of the areas.
It is the issues that have been raised in Portland, in
Atlanta, in Chicago and Seattle, and then also to a lesser
extent in Southern California, primarily L.A. and Orange
County, and then by New York City area stakeholders where we
have heard concerns regarding facial recognition technology.
One of our primary roles is to be the eyes and ears of the
Department, and we inform our colleagues at CBP, at DHS S&T, at
OBIM. Here are the concerns that we are seeing. How do we work
together to try and address some of these problems or potential
problems before they have even greater effect?
Then also, on the back end we have a robust compliance
process, and while we do not have an active investigation right
now on facial recognition, that is always something that we are
looking at. If we see a trend, we will most certainly open an
investigation and, again, advice in that way as well.
Ms. Jackson Lee. Would you be kind enough, Mr. Chairman, to
let Mr. Romine answer his question?
As he answers, Mr. Chairman, I just want to say this on the
record, if we can get answers from Mr. Wagner about what is
stored in terms of retaining information.
Chairman Thompson. Ms. Jackson Lee.
Ms. Jackson Lee. Thank you. Mr.----
Chairman Thompson. No. Dr. Romine, you can answer the
question. You can submit in writing to Ms. Jackson Lee.
Ms. Jackson Lee. No, that is what I am saying.
Chairman Thompson. Yes.
Ms. Jackson Lee. Yes. Thank you.
Chairman Thompson. Be happy to do it.
Ms. Jackson Lee. Thank you.
Mr. Romine, my question was----
Mr. Romine. I beg your pardon, ma'am?
Ms. Jackson Lee. Yes. My question was: What are you doing
to be accurate in your testing?
You said you did not know whether you had the accurate app
that they were using. What are you doing to be aggressive in
making sure that we do not have the bias in these algorithms?
Mr. Romine. Yes, ma'am. The tests that we undertake are
intended to determine whether there are demographic
differences, commonly called bias. The fact that I know there
is strong interest in testing with data that is more
representative, and we have signed a recent MOU with the CBP to
undertake continued testing to make sure that we are doing the
very best that we can to provide the information that they need
to make sound decisions.
Ms. Jackson Lee. Thank you very much.
I yield back. Thank you.
Chairman Thompson. Thank you.
The Chair recognizes the gentleman from Texas, Mr. Green.
Mr. Green of Texas. Thank you, Mr. Chairman.
I thank the witnesses for appearing as well.
I would like to address some intelligence that has been
afforded me. The indication is that NIST found that Asian and
African American faces were 10 times more likely, well, 10 to
100 times more likely to be misidentified than white faces.
I am curious as to whether or not there is something
inherent in the technology that creates an inverse relationship
with reference to the identification of whites juxtaposed to
African Americans and Asians.
Is there something inherent in the technology, meaning if
you want to absolutely identify whites, will there be something
that you cannot adjust such that you will get the same absolute
identification with minorities, Asians, African Americans?
Or if you want to absolutely identify African Americans and
Asians, will you, as a result of technology, not be able to
properly identify whites?
Mr. Romine. It is a very interesting question, Congressman.
Mr. Green of Texas. Thank you.
Mr. Romine. Let me clarify first that those differentials
that we observed were not in the case of identification but
rather verification, the one-to-one testing rather than the
one-to-many testing. In general, we saw those demographic
differences for African Americans and for Pacific Islanders and
Asians as well.
But in the case that you are talking about, our work has
not to-date focused on cause and effect. What is it that is
causing the algorithms to exhibit certain kinds of behavior? We
are really just testing the performance.
So I do not know the answer to your question.
Mr. Green of Texas. My question was interesting, as you put
it. Your answer is intriguing because this is not the first
opportunity for the word to be heard that we have these
difficulties, and at some point, it would seem that we would
move from testing technology as it is to understanding why
technology performs the way it does.
Help me to understand why we have not made that move?
Mr. Romine. The question that you asked is a very
challenging open research question, but we do have some
indications.
There are algorithms that have been submitted to our
testing from Asian countries that do not exhibit the
demographic differentials on Asian faces. So we cannot
guarantee, but we think that is an indication that the training
data that are being used for the algorithm development may have
a significant impact on their ability to discern or exhibit
demographic differences for different populations.
Mr. Green of Texas. Do you believe that it is important for
us to move expeditiously to answer this question, to resolve
this issue such that we do not find ourselves having deployed
something en masse that we know to be defective or have some
degree of inefficiencies associated with it?
The efficacy of this is important.
Mr. Romine. Yes, sir, and I think those are two different
things to think about. The performance testing that we
currently execute can help operational agencies ensure that
they are not deploying things that exhibit demographic
differentials.
The research question that you teed up that is fascinating
about what are the causes of these demographic differentials is
a much deeper question and much more difficult, I think.
Mr. Green of Texas. Well, is it fair to say that the
countries--and I have about 45 seconds left--but the countries
that employ the technology that have indicated to you they are
having fewer challenges, is it fair to say that that technology
also captures white men sufficiently?
Mr. Romine. In the testing that we did for the specific one
that I am referring to, the high-performing algorithms from
Asian countries that do not exhibit the demographic differences
on Asians, it is in comparison to Caucasian faces that I made
that statement.
So there is no difference in the performance or discernable
difference in the performance on Caucasian faces and Asian
faces from certain Asian-developed algorithms, and one
speculation is that it may be the training data that are used.
Mr. Green of Texas. Thank you, Mr. Chairman. I yield back.
Chairman Thompson. Thank you very much.
The Chair recognizes the gentleman from Rhode Island for 5
minutes. Mr. Langevin.
Mr. Langevin. Thank you, Mr. Chairman.
I want to thank our witnesses for your testimony here
today, and thank you for you are doing, your dedication to this
issue, a very important issue.
I certainly believe that technology is an important part of
the solutions that some of our most vexing issues and
challenges, including how to manage an ever-growing number of
international travelers. So it has been a good discussion here
today.
What I wanted to ask of either Mr. Wagner or Mr. Mina, we
know that in technological solutions, such as facial
recognition software, the algorithms are only as good as the
data that inform them. So I want to know how has CBP adjusted
or augmented the data that it uses to train its facial
recognition software.
What are you doing to ensure the software is continually
updated as more robust data sets and algorithms are
incorporated into training?
Mr. Wagner. That is where we work closely with the vendor,
whose algorithm we are using, NEC, and we work closely with
them to incorporate their updates and their latest and greatest
products into how we are using them.
Then as we review the data, you know, we look to make those
operational adjustments, which do impact metrics, and again,
that is going to be the quality of the photograph, the quality
of the camera, the human factors.
The size of the gallery is really important, and you know,
in this, it tested galleries up to, I think, 12 million people.
You know, on the margins of the capabilities of these
algorithms, we are doing this on a couple thousand, and
interesting correlations are how much better improved is your
match rates and what is the impact on any potential demographic
biases on a much smaller gallery or sample size.
I think that is what we were getting at earlier, that what
are these variables that we can raise or lower to help address
some of what the error rates are showing us.
Mr. Langevin. OK. So to that point then, how does CBP
incorporate feedback from officers about errors that facial
recognition software has made in the field?
Because the machine, it learns. When the officer is
looking, interacting with someone, and the software does not
get it correct, unless that feedback is fed back into the
system, the system does not learn.
Mr. Wagner. Oh, absolutely, and that is where we look at
the system logs themselves, but we also talk to the officers.
They provide the feedback, and then we are also on-site to
witness and observe and discuss with those officers as we
deploy these.
Mr. Langevin. That is important.
So I understand the Trusted Traveler Program shares
information with other countries, and how does CBP share
biometric information with other countries and what steps does
it take to ensure that those countries use the data
responsibly?
Is that accurate, No. 1, what my understanding is?
How are we guarding that data to make sure that they are
protected it?
Mr. Wagner. I am trying to think of when. I am not aware of
how we would share or if we are even sharing.
Mr. Langevin. With the Trusted Traveler Program.
Mr. Wagner. We do not share. We might share a person's
status that they are approved in the program, but we are not
actually sharing, say, their fingerprints.
Mr. Langevin. OK. So let me ask that one for the record,
and I would ask that you get back to me on that.
Mr. Wagner. Yes.
Mr. Langevin. This is important.
What types of information do we share under the Trusted
Traveler Program? I think that is important for us to know.
If we do share, whatever information we share, I want to
know what steps we take to ensure that those countries use that
data responsibly.
So I know that this question has been touched on earlier.
So I am going to ask it perhaps in a different way, but just
prior to our hearing on this topic, last July we were notified
of a cyber incident on the network of a CBP subcontractor.
Someone claiming to be a foreign agent gained access to tens of
thousands of photos of driver's faces and license plates at a
port of entry along the Southern Border.
How is CBP ensuring that the personal data it collects for
facial recognition technology screening programs, whether by
the Government directly or by vendors or their private-sector
partners, are being protected from inadvertent or otherwise
unauthorized access?
Also, what assurances can you give our committee that the
root causes of the May 2019 breach have been addressed so as to
reduce the likelihood of another breach?
Mr. Wagner. So the airlines and airports that provide the
cameras that take the pictures to transmit them to CBP, we have
a signed set of business requirements with them which they
commit to not storing, not sharing, not saving any of the
photographs that they take.
They take the picture, have to transmit it to us, and purge
it from their system.
One of the other conditions is that they have to be
available for CBP to audit their cameras and their technology
to ensure that they are following those rules.
We are about to commence an audit on one of the airlines in
the next couple of months and start that process to do that,
but to make sure that that is not happening.
Mr. Langevin. Thank you.
Mr. Chairman, I would ask that Mr. Wagner get back with me
in writing as soon as possible on that Trusted Traveler Program
and what information is shared with partners.
Chairman Thompson. OK.
Mr. Langevin. Thank you, Mr. Chairman.
Chairman Thompson. Thank you.
Did the gentlelady from Texas want to ask a question since
everybody else has asked theirs?
Ms. Jackson Lee. Yes, Mr. Chairman. Thank you so very much.
First of all, I will ask unanimous consent to place in the
record, not to the witnesses, but the headline reads, ``Amazon
Facial Recognition Mistakenly Confused 28 Congressmen with
Known Criminals,'' July 26, 2018.
Chairman Thompson. Without objection.
[The information follows:]
Article Submitted by Honorable Sheila Jackson Lee
Amazon facial recognition mistakenly confused 28 Congressmen with known
criminals
By Sean Hollister, July 26, 2018, 12:45 PM PDT
https://www.cnet.com/news/amazon-facial-recognition-thinks-28-
congressmen-look-like-known-criminals-at-default-settings/
The ACLU says it's evidence that Congress should step in. Amazon
says the ACLU didn't test properly.
Amazon is trying to sell its Rekognition facial recognition
technology to law enforcement, but the American Civil Liberties Union
doesn't think that's a very good idea. And today, the ACLU provided
some seemingly compelling evidence--by using Amazon's own tool to
compare 25,000 criminal mugshots to Members of Congress.
Sure enough, Amazon's tool thought 28 different Members of Congress
looked like people who've been arrested.
Here's the full list, according to the ACLU:
Senate:
Johnny Isakson (R-Georgia)
Ed Markey (D-Massachusetts)
Pat Roberts (R-Kansas)
House:
Sanford Bishop (D-Georgia)
G.K. Butterfield (D-North Carolina)
Lacy Clay (D-Missouri)
Mark DeSaulnier (D-California)
Adriano Espaillat (D-New York)
Ruben Gallego (D-Arizona)
Tom Garrett (R-Virginia)
Greg Gianforte (R-Montana)
Jimmy Gomez (D-California)
Raul Grijalva (D-Arizona)
Luis Gutierrez (D-Illinois)
Steve Knight (R-California)
Leonard Lance (R-New Jersey)
John Lewis (D-Georgia)
Frank LoBiondo (R-New Jersey)
Dave Loebsack (D-Iowa)
David McKinley (R-West Virginia)
John Moolenaar (R-Michigan)
Tom Reed (R-New York)
Bobby Rush (D-Illinois)
Norma Torres (D-California)
Marc Veasey (D-Texas)
Brad Wenstrup (R-Ohio)
Steve Womack (R-Arkansas)
Lee Zeldin (R-New York)
That's a lot of Congresspeople who may soon have some very valid
questions about facial recognition and its potential to be abused--
particularly since Amazon thinks the ACLU didn't use it properly to
begin with! It turns out that the ACLU got its mugshot matches by using
the Rekognition software at its default 80-percent confidence threshold
setting, rather than the 95-percent-plus confidence level that Amazon
recommends for law enforcement agencies.
It turns out that the ACLU got its mugshot matches by using the
Rekognition software at its default 80-percent confidence threshold
setting, rather than the 95-percent plus confidence level that Amazon
recommends for law enforcement agencies.
``While 80 percent confidence is an acceptable threshold for photos
of hot dogs, chairs, animals, or other social media use cases, it
wouldn't be appropriate for identifying individuals with a reasonable
level of certainty. When using facial recognition for law enforcement
activities, we guide customers to set a threshold of at least 95
percent or higher,'' an Amazon spokesperson told CNET by email.
But an ACLU lawyer tells CNET that Amazon doesn't necessarily steer
law enforcement agencies toward that higher threshold--if a police
department uses the software, it'll be set to the same 80-percent
threshold by default and won't ask them to change it even if they
intend to use it to identify criminals. ``Amazon makes no effort to ask
users what they are using Rekognition for,'' says ACLU attorney Jacob
Snow.
A source close to the matter tells CNET that when Amazon works with
law enforcement agencies directly, like the Orlando Police Department,
it teaches them how to reduce false positives and avoid human bias. But
there's nothing to necessarily keep other agencies from simply using
the tool the same way the ACLU did, instead of developing a
relationship with Amazon.
It's worth noting that false positives are (currently!) an accepted
part of facial recognition technology. Nobody--including the ACLU--is
saying police would arrest someone based on a false match alone. Facial
recognition narrows down the list of suspects, and then humans take
over. Recently, facial recognition helped ID the Russian assassins who
poisoned a spy in the UK, as well as the Capital Gazette shooter.
And Amazon didn't actually create that many false positives even at
the 80 percent confidence ratio, compared to, say, the UK Metropolitan
Police's facial recognition tech.
But the ACLU worries that Amazon's false positives might bias a
police officer or government agent to search, question, or potentially
draw a weapon when they shouldn't--and we've all seen how those
encounters can turn deadly. And the ACLU notes that Amazon's tech seems
to have over-represented people of color.
Should Congress regulate facial recognition? Microsoft thinks so,
and now 28 Members of Congress have some very personal food for
thought--95-percent confidence threshold or no.
In the hours since the ACLU's test was brought to light, five
Congressmen have sent letters to Amazon CEO Jeff Bezos asking for
answers and an immediate meeting. You can read the letters here.
Ms. Jackson Lee. So, Mr. Wagner, I just wanted to ask you.
Are you using Amazon technology?
Mr. Wagner. We are not using their matching algorithm.
Ms. Jackson Lee. Thank you, Mr. Chairman.
My question is you gave Congressman Langevin sort-of a
detailed response. So let me try to change it around to: Do you
have a team that is directly responsible not just for the
implementation, but for the internal analysis of the
utilization of the app or the technology that you are using so
that it is on-site, so you are able to get first-hand knowledge
of the violations or let me use the word ``abuses'' by way of
the technology?
Is that information coming back to your office? When I say
that, to your sector.
Mr. Wagner. Yes. Part of it is our office that does it, and
then working in conjunction with our field locations.
Ms. Jackson Lee. So do you have a team that is just
responding to that, if you would?
Mr. Wagner. We have teams that review the data, review the
reports, review the functioning of the systems, review the
compliance of the officers using the technology, yes.
Ms. Jackson Lee. Mr. Chairman, I will just say this. I know
we have a lot of work. You have a lot of work. You do a lot,
but maybe there could be a Classified briefing.
I would like to do a deeper dive on how that is done and
how that is kept and whether they store, how long they keep the
data on Mr. Jones or Mr. Aman and Mr. various named persons,
how long they keep the data.
Chairman Thompson. Well, we will work through it.
Mr. Wagner. The data is all stored in compliance with the
Systems of Record Notices of where that data is stored. So the
photograph of the U.S. citizen that we take is only stored for
12 hours and then purged.
A picture of a foreign national is sent over to IDENT, the
Department's database, where it is stored for 75 years.
The record of the border crossing, the biographical
information is then stored in other systems attributable to the
System of Record Notices attributable to those.
Ms. Jackson Lee. Very interesting. Thank you.
Chairman Thompson. We will follow up on your request.
Ms. Jackson Lee. Thank you, witnesses. Thank you.
Chairman Thompson. Let me insert in the record a letter
from the Electronic Privacy Information Center and a press
release from the U.S. Travel Association.
[The information follows:]
Letter From the Electronic Privacy Information Center
February 5, 2020.
The Honorable Bennie G. Thompson, Chairman,
The Honorable Mike Rogers, Ranking Member,
Committee on Homeland Security, U.S. House of Representatives, H2-176
Ford House Office Building, Washington, DC 20515.
Dear Chairman Thompson and Ranking Member Rogers: We write to you
in advance of the hearing on ``About Face: Examining the Department of
Homeland Security's Use of Facial Recognition and Other Biometric
Technologies, Part II.''\1\ EPIC supports a moratorium on facial
recognition technology for mass surveillance. This committee should
halt DHS's use of face surveillance technology.
---------------------------------------------------------------------------
\1\ About Face: Examining the Department of Homeland Security's Use
of Facial Recognition and Other Biometric Technologies, Part II, House
Comm. on Homeland Security, 116th Cong. (Feb. 6, 2020), https://
homeland.house.gov/activities/hearings/about-face-examining-the-
department-of-homeland-securitys-use-of-facial-recognition-and-other-
biometric-technologies-part-ii.
---------------------------------------------------------------------------
The Electronic Privacy Information Center (``EPIC'') is a public
interest research center established in 1994 to focus public attention
on emerging privacy and civil liberties issues.\2\ EPIC is focused on
protecting individual privacy rights, and we are particularly
interested in the privacy problems associated with surveillance.\3\
Last year, EPIC filed a lawsuit against the Customs and Border
Protection (``CBP'') agency for failure to establish necessary privacy
safeguards for the collection of facial images at U.S. borders.\4\
---------------------------------------------------------------------------
\2\ See About EPIC, EPIC.org, https://epic.org/epic/about.html.
\3\ EPIC, EPIC Domestic Surveillance Project, https://epic.org/
privacy/surveillance/.
\4\ EPIC v. U.S. Customs and Border Protection, No. 19-cv-689
(D.D.C. filed Mar. 12, 2019); See https://epic.org/foia/dhs/cbp/alt-
screening-procedures/.
---------------------------------------------------------------------------
a call to ban face surveillance
EPIC and the Public Voice Coalition are leading a global campaign
to establish a moratorium on ``face surveillance,'' the use of facial
recognition for mass surveillance.\5\ In October 2019 more than 100
NGO's and hundreds of experts endorsed our petition.\6\ The signatories
stated:
---------------------------------------------------------------------------
\5\ EPIC, Ban Face Surveillance, https://epic.org/
banfacesurveillance/.
\6\ The Public Voice, Declaration: A Moratorium on Facial
Recognition Technology for Mass Surveillance Endorsements, https://
thepublicvoice.org/ban-facial-recognition/endorsement/.
---------------------------------------------------------------------------
We urge countries to suspend the further deployment of
facial recognition technology for mass surveillance;
We urge countries to review all facial recognition systems
to determine whether personal data was obtained lawfully and to
destroy data that was obtained unlawfully;
We urge countries to undertake research to assess bias,
privacy and data protection, risk, and cyber vulnerability, as
well as the ethical, legal, and social implications associated
with the deployment of facial recognition technologies; and
We urge countries to establish the legal rules, technical
standards, and ethical guidelines necessary to safeguard
fundamental rights and comply with legal obligations before
further deployment of this technology occurs.
Courts and regulators are also listening. There is growing
awareness of the need to bring this technology to a halt. The State of
California prohibited the use facial recognition on police-worn body
cameras. Several cities in the U.S. have banned the use of facial
recognition systems, and there is a growing protest around the world.
For example, In 2019 the Swedish Data Protection Authority prohibited
the use of facial recognition in schools. EPIC has published a resource
of laws, regulations, legal decisions and reports on face surveillance
worldwide at https://epic.org/banfacesurveillance/.
threats to privacy and civil liberties
Facial recognition poses serious threats to privacy and civil
liberties and can be deployed covertly, remotely, and on a mass scale.
There is a lack of well-defined regulations controlling the collection,
use, dissemination, and retention of biometric identifiers. Ubiquitous
identification by commercial or Government entities eliminates the
individual's ability to control the disclosure of their identities,
creates new opportunities for tracking and monitoring, increases the
security risks from data breaches. An individual's ability to control
disclosure of his or her identity is an essential aspect of personal
freedom and autonomy. The use of facial recognition erodes these
freedoms.
There is little a person in the United States could do to prevent
the capture of their image by the Government or a private company if
face surveillance is deployed. Participation in society necessarily
requires participation in public spaces. But ubiquitous and near
effortless identification eliminates the individual's ability to
control the disclosure of their identities to others. Strangers will
know our identities as readily as our friends and family members.
use of face surveillance in china
Face surveillance capabilities have been on full display in China.
China is not only the leading government for face surveillance
technology, it is also the leading exporter of the technology.\7\ The
Chinese government has implemented a massive facial recognition
surveillance system.\8\ China has leveraged its surveillance network to
implement an ``advanced facial recognition technology to track and
control the Uighurs, a largely Muslim minority.''\9\ And China
continues to expand the use of facial recognition technology. A
university in China is testing the use of facial recognition to monitor
whether students attend classes and to track their attention during
lectures.\10\ To register a new mobile phone number in China now
requires one to submit to a facial scan.\11\ Trials have also begun to
use facial recognition at security checkpoints in the subway
system.\12\
---------------------------------------------------------------------------
\7\ Steven Feldstein, The Global Expansion of AI Surveillance 13-15
(Sept. 2019), https://carnegieendowment.org/files/WP-Feldstein-
AISurveillance_final1.pdf.
\8\ Simon Denyer, China's Watchful Eye, Wash. Post (Jan. 7, 2018),
https://www.washingtonpost.com/news/world/wp/2018/01/07/feature/in-
china-facial-recognition-is-sharp-end-of-a-drive-for-total-
surveillance/.
\9\ Paul Mozur, One Month, 500,000 Face Scans: How China is Using
A.I. to Profile a Minority, N.Y. Times (Apr. 14, 2019), https://
www.nytimes.com/2019/04/14/technology/china-surveillance-artificial-
intelligence-racial-profiling.html.
\10\ Brendan Cole, Chinese University Tests Facial Recognition
System to Monitor Attendance and Students' Attention to Lectures,
Newsweek (Sept. 2, 2019), https://www.newsweek.com/nanjing-china-
facial-recognition-1457193.
\11\ Kyle Wiggers, AI Weekly: In China, You Can No Longer Buy a
Smartphone without a Face Scan, VentureBeat (Oct. 11, 2019), https://
venturebeat.com/2019/10/11/ai-weekly-in-china-you-can-no-longer-buy-a-
smartphone-without-a-face-scan/.
\12\ Wan Lin, Beijing Subway Station Trials Facial Recognition,
Global Times (Dec. 1, 2019), http://www.globaltimes.cn/content/
1171888.shtml.
---------------------------------------------------------------------------
In Hong Kong, where protests have been on-going since March, face
scans have become a weapon. Protesters fear that facial recognition
technology is being used to identify and track them.\13\ In response to
this fear, protesters have resorted to covering their faces and have
taken down facial recognition cameras. Hong Kong reacted by banning
masks and face paint.\14\ Many of the demonstrators worry that the mass
surveillance implemented on the mainland of China will be implemented
in Hong Kong.
---------------------------------------------------------------------------
\13\ Paul Mozur, In Hong Kong Protests, Faces Become Weapons, N.Y.
Times (July 26, 2019), https://www.nytimes.com/2019/07/26/technology/
hong-kong-protests-facial-recognition-surveillance.html.
\14\ Matt Novak, Hong Kong Announces Ban on Masks and Face Paint
That Helps Protesters Evade Facial Recognition, Gizmodo (Oct. 4, 2019),
https://gizmodo.com/hong-kong-announces-ban-on-masks-and-face-paint-
that-he-1838765030.
---------------------------------------------------------------------------
face surveillance in the united states
The implementation of facial recognition technology by Government
and commercial actors in the United States is pushing the U.S. toward a
similar mass surveillance infrastructure. Already some schools are
implementing the use of facial recognition technology.\15\ Customs and
Border Protection (CBP) is using facial recognition on travelers
entering and exiting the U.S.\16\ And airlines are using CBP's facial
recognition system to conduct flight check-ins, check bags, and board
flights.\17\ The Rochester airport has implemented the surveillance
infrastructure to perform facial recognition on every person that
enters the airport.\18\ Amazon drafted plans to use their Ring
surveillance cameras to create neighborhood watch lists that leverage
facial recognition.\19\ Retailers have implemented the use of facial
recognition at their stores.\20\ A landlord in Brooklyn wanted to use
facial recognition as the means to gain entry into a rent-stabilized
apartment building.\21\ Facial recognition is being used at major
sporting events \22\ and concerts.\23\ And the companies that are
creating the facial recognition algorithms are often using--without
consent--millions of photos scraped from social media sites and other
webpages in order train the algorithms.\24\
---------------------------------------------------------------------------
\15\ Tom Simonite and Gregory Barber, The Delicate Ethics of Using
Facial Recognition in Schools, Wired (Oct. 17, 2019), https://
www.wired.com/story/delicate-ethics-facial-recognition-schools/.
\16\ Davey Alba, The US Government Will Be Scanning Your Face At 20
Top Airports, Documents Show, BuzzFeed (Mar. 11, 2019), https://
www.buzzfeednews.com/article/daveyalba/these-documents-reveal-the-
governments-detailed-plan-for?ref=bfnsplash.
\17\ See, e.g., Kathryn Steele, Delta Unveils First Biometric
Terminal in U.S. in Atlanta; next stop: Detroit, Delta News Hub,
https://news.delta.com/delta-unveils-first-biometric-terminal-us-
atlanta-next-stop-detroit.
\18\ James Gilbert, Facial Recognition Heading to Rochester Airport
Despite Concerns, Rochester First (June 26, 2019), https://
www.rochesterfirst.com/news/local-news/facial-recognition-heading-to-
airport-despite-concerns/.
\19\ Sam Biddle, Amazon's Ring Planned Neighborhood ``Watch Lists
`` Built on Facial Recognition, The Intercept (Nov. 26, 2019), https://
theintercept.com/2019/11/26/amazon-ring-home-security-facial-
recognition/.
\20\ Nick Tabor, Smile! The Secretive Business of Facial-
Recognition Software in Retails Stores, New York Intelligencer (Oct.
20, 2018), http://nymag.com/intelligencer/2018/10/retailers-are-using-
facial-recognition-technology-too.html.
\21\ Gina Bellafante, The Landlord Wants Facial Recognition in Its
Rent-Stabilized Buildings. Why?, N.Y. Times (Mar. 28, 2019), https://
www.nytimes.com/2019/03/28/nyregion/rent-stabilized-buildings-facial-
recognition.html.
\22\ Ryan Rodenberg, Sports Betting and Big Brother: Rise of Facial
Recognition Cameras, ESPN (Oct. 3, 2018 ), https://www.espn.com/chalk/
story/_/id/24884024/why-use-facial-recognition-cameras-sporting-events-
the-rise.
\23\ Steve Knopper, Why Taylor Swift Is Using Facial Recognition at
Concerts, Rolling Stone (Dec. 13, 2018), https://www.rollingstone.corn/
music/music-news/taylor-swift-facial-recognition-concerts-768741/.
\24\ Kashmir Hill, The Secretive Company That Might End Privacy as
We Know It, N.Y. Times (Jan. 18, 2020), https://www.nytimes.com/2020/
01/18/technology/clearview-privacy-facial-recognition.html.
---------------------------------------------------------------------------
It is important to note that not all uses of facial recognition are
equally problematic. For instance, where the user has control and there
is no Government mandate, such as using Face ID for iPhone
authentication, the same privacy issues do not arise. Facial
recognition can also be used for verification or authentication using
1:1 matching--that is, where the system does not check every record in
a database for a match, but matches the individual's face to their
claimed identity.\25\ This 1:1 matching is a much more privacy
protective implementation of facial recognition. 1:1 matching does not
require a massive biometric database, there is no need to retain the
image, and the machines conducting the 1:1 match do not need to be
connected to the cloud. Such an implementation virtually eliminates
data breach risks and the chance of mission creep.
---------------------------------------------------------------------------
\25\ Lucas D. lntrona and Helen Nissenbaum, Facial Recognition
Technology: A Survey of Policy and Implementation Issues, Ctr. for
Catastrophe Preparedness & Response, N.Y. Univ., 11 (2009), available
at https://nissenbaum.tech.cornell.edu/papers/
facial_recognition_report.pdf.
---------------------------------------------------------------------------
face surveillance in airports
Recently, new privacy risks have arisen with the deployment of
facial recognition technology at U.S. airports following a 2017
Executive Order to ``expedite the completion and implementation of
biometric entry exit tracking system.''\26\ Customs and Border
Protection (``CBP'') has now implemented the Biometric Entry-Exit
program for international travelers at 17 airports.\27\ TSA is quickly
moving to leverage CBP's Biometric Entry-Exit program to expand the use
of facial recognition at airports.\28\
---------------------------------------------------------------------------
\26\ Exec. Order No. 13,780 8.
\27\ Davey Alba, The US Government Will Be Scanning Your Face At 20
Top Airports, Documents Show (Mar. 11, 2019), https://
www.buzzfeednews.corn/article/daveyalba/these-documents-reveal-the-
governments-detailed-plan-for.
\28\ TSA, TSA Biometrics Roadmap (Sept. 2018), https://www.tsa.gov/
sites/default/files/tsa_biometrics_roadmap.pdf.
---------------------------------------------------------------------------
TSA has already deployed facial recognition technology at two TSA
Checkpoints.\29\ In September 2018, TSA released a ``TSA Biometrics
Roadmap,'' detailing its plans to use facial recognition, including on
domestic travelers.\30\ The Roadmap makes clears TSA's intention to
leverage CBP's facial recognition capabilities implemented as part of
the Biometric Entry-Exit Program. But corresponding privacy safeguards
have not yet been established.
---------------------------------------------------------------------------
\29\ Trans. Security Admin., Travel Document Checker Automation
Using Facial Recognition, (Aug. 2019), https://www.dhs.gov/publication/
dhstsapia-046-travel-document-checker-automation-using-facial-
recognition; U.S. Customs and Border Protection, CBP Deploys Facial
Recognition Biometric Technology at 1 TSA Checkpoint at JFK Airport
(Oct. 11, 2017), https://www.cbp.gov/newsroom/national-media-release/
cbp-deploys-facial-recognition-biometric-techno- logy-1-tsa-checkpoint.
\30\ TSA, TSA Biometrics Roadmap (Sept. 2018), https ://
www.tsa.gov/sites/default/files/tsa_biometrics_roadmap.pdf.
---------------------------------------------------------------------------
In response to EPIC's Freedom of Information Act request, CBP
recently released 346 pages of documents detailing the agency's
scramble to implement the flawed Biometric Entry-Exit system, a system
that employs facial recognition technology on travelers entering and
exiting the country. The documents obtained by EPIC describe the
administration's plan to extend the faulty pilot program to major U.S.
airports. The documents obtained by EPIC were covered in-depth by
Buzzfeed.\31\
---------------------------------------------------------------------------
\31\ Davey Alba, The US Government Will Be Scanning Your Face At 20
Top Airports, Documents Show (Mar. 11, 2019), https://
www.buzzfeednews.com/article/daveyalba/these-documents-reveal-the-
govemments-detailed-plan-for.
---------------------------------------------------------------------------
Based on the documents obtained, EPIC determined that there are few
limits on how airlines will use the facial recognition data collected
at airports.\32\ Only recently has CBP changed course and indicated
that the agency will require airlines to delete the photos they take
for the Biometric Entry-Exit program.\33\ No such commitment has been
made by TSA. Indeed, TSA's Roadmap indicates that the agency wants to
expand the dissemination of biometric data as much as possible,
stating:
---------------------------------------------------------------------------
\32\ See CBP Memorandum of Understanding Regarding Biometric Pilot
Project, https://epic.org/foia/dhs/cbp/biometric-entry-exit/MOU-
Biometric-Pilot-Project.pdf.
\33\ Ashley Ortiz, CBP Program and Management Analyst, Presentation
before the Data Privacy & Integrity Advisory Committee, slide 23 (Dec.
2018), https://www.dhs.gov/sites/default/files/publications/SLIDES-
DPIAC-Public%20Meeting%2012%2010-2018.pdf.
``TSA will pursue a system architecture that promotes data sharing to
maximize biometric adoption throughout the passenger base and across
the aviation security touch points of the passenger experience.''\34\
---------------------------------------------------------------------------
\34\ TSA, TSA Biometrics Roadmap, 17 (Sept. 2018).
TSA seeks to broadly implement facial recognition through ``public-
private partnerships'' in an effort to create a ``biometrically-enabled
curb-to-gate passenger experience.''\35\ Currently, TSA plans to
implement an opt-in model of facial recognition use for domestic
travelers but there are no guarantees that in the future TSA will not
require passengers to participate in facial recognition or make the
alternative so cumbersome as to essentially require passengers to opt-
in.
---------------------------------------------------------------------------
\35\ Id. at 19.
---------------------------------------------------------------------------
Preserving the ability of U.S. citizens to forgo facial recognition
for alternative processes is one of the privacy issues with CBP's
Biometric Entry-Exit program. Senator Markey (D-MA) and Senator Lee (R-
UT) called for the CBP to suspend facial recognition at the border to
ensure that travelers are able to opt out of facial recognition if they
wish.\36\
---------------------------------------------------------------------------
\36\ Press Release, Sens. Edward Markey and Mike Lee, Senators
Markey and Lee Call for Transparency on DHS Use of Facial Recognition
Technology (Mar. 12, 2019), https://www.markey.senate.gov/news/press-
releases/senators-markey-and-lee-call-for-transparency-on-dhs-use-of-
facial-recognition-technology.
---------------------------------------------------------------------------
In fact, EPIC recently sued CBP for all records related to the
creation and modification of alternative screening procedures for the
Biometric Entry-Exit program.\37\ The alternative screening procedure
for U.S. travelers that opt out of facial recognition should be a
manual check of the traveler's identification documents. CBP, however,
has provided vague and inconsistent descriptions of alternative
screening procedures in both its ``Biometric Exit Frequently Asked
Questions (FAQ)'' webpage \38\ and the agency's privacy impact
assessments.\39\ The creation and modification of CBP's alternative
screening procedures underscores CBP's unchecked ability to modify
alternative screening procedures while travelers remain in the dark
about how to protect their biometric data.
---------------------------------------------------------------------------
\37\ EPIC v. CBP, 19-cv-00689, Complaint, https://epic.org/foia/
cbp/altemative-screening-procedures/1-Complaint.pdf.
\38\ CBP, Biometric Exit Frequently Asked Questions (FAQs), https:/
/www.cbp.gov/travel/biometrics/biometric-exit-faqs.
\39\ U.S. Dep't of Homeland Sec., DHS/CBP/PIA-030(b), Privacy
Impact Assessment Update for the Traveler Verification Service (TVS):
Partner Process 8 (2017), https://www.dhs.gov/sites/default/files/
publications/privacy-pia-cbp030-tvs-may2017.pdf; see also U.S. Dep't of
Homeland Sec., DHS/CBP/PIA-030(c), Privacy Impact Assessment Update for
the Traveler Verification Service (TVS): Partner Process 5-6 (2017),
https://www.dhs.gov/sites/default/files/publications/privacy-pia-
cbp030-tvs-appendixb-july2018.pdf; U.S. Dep't of Homeland Sec., DHS/
CBP/PIA-056, Privacy Impact Assessment for the Traveler Verification
Service 2 (2018), https://www.dhs.gov/sites/default/files/publications/
privacy-pia-cbp-0-tvs-november2018_2.pdf.
---------------------------------------------------------------------------
face surveillance and ai
It is also becoming increasingly clear that AI tools are being
deployed with facial recognition to accelerate the deployment of
technology not only for identification but also for scoring. As we
explained recently in the New York Times, ``The United States must work
with other democratic countries to establish red lines for certain AI
applications and ensure fairness, accountability, and transparency as
AI systems are deployed.''\40\ In a subsequent letter to the New York
Times, we warned of the growing risk of the Chinese AI model, and
specifically explained, ``China also dominates the standards-setting
process for techniques like facial recognition.''\41\
---------------------------------------------------------------------------
\40\ Marc Rotenberg, The Battle Over Artificial Intelligence, N.Y.
Times, Apr. 18, 2019, https://www.nytimes.com/2019/04/18/opinion/
letters/artificial-intelligence.html. In the introduction to the EPIC
AI Policy Sourceboook and in a subsequent letter to the New York Times,
we warned of the growing risk of the Chinese AI model. Marc Rotenberg
and Len Kennedy, Surveillance in China: Implications for Americans,
N.Y. Times, Dec. 19, 2019, https://www.nytimes.com/2019/12/19/opinion/
letters/surveillance-china.html.
\41\ Marc Rotenberg and Len Kennedy, Surveillance in China:
Implications for Americans, N.Y. Times, Dec. 19, 2019, https://
www.nytimes.com/2019/12/19/opinion/letters/surveillance-china.html.
---------------------------------------------------------------------------
Society is simply not in a place right now for the wide-scale
deployment of facial recognition technology. It would be a mistake to
deploy facial recognition at this time. We urge the committee to
support a ban of DHS's further deployment of face surveillance
technology.
We ask that this statement be entered in the hearing record. EPIC
looks forward to working with the committee on these issues of vital
importance to the American public.
Sincerely,
Marc Rotenberg,
EPIC President.
Caitriona Fitzgerald,
EPIC Policy Director.
Jeramie Scott,
EPIC Senior Counsel.
Attachment.--Declaration: A Moratorium on Facial Recognition Technology
for Mass Surveillance, The Public Voice, Tirana Albania (October 2019).
______
Attachment.--Declaration: A Moratorium on Facial Recognition Technology
for Mass Surveillance
October 2019, Tirana, Albania
We the undersigned call for a moratorium on the use of facial
recognition technology that enables mass surveillance.
We recognize the increasing use of this technology for commercial
services, Government administration, and policing functions. But the
technology has evolved from a collection of niche systems to a powerful
integrated network capable of mass surveillance and political control.
Facial recognition is now deployed for human identification,
behavioral assessment, and predictive analysis.
Unlike other forms of biometric technology, facial recognition is
capable of scrutinizing entire urban areas, capturing the identities of
tens or hundreds of thousands of people at any one time.
Facial recognition can amplify identification asymmetry as it tends
to be invisible or at best, opaque.
Facial recognition can be deployed in almost every dimension of
life, from banking and commerce to transportation and communications.
We acknowledge that some facial recognition techniques enable
authentication for the benefit of the user. However facial recognition
also enables the development of semi-autonomous processes that minimize
the roles of humans in decision making.
We note with alarm recent reports about bias, coercion, and fraud
in the collection of facial images and the use of facial recognition
techniques. Images are collected and used with forced consent or
without consent at all.
We recall that in the 2009 Madrid Declaration, civil society called
for a moratorium on the development or implementation of facial
recognition, subject to a full and transparent evaluation by
independent authorities and through democratic debate.
There is growing awareness of the need for a moratorium. In 2019
the Swedish Data Protection Authority prohibited the use of facial
recognition in schools. The State of California prohibited the use
facial recognition on police-worn body cameras. Several cities in the
United States have banned the use of facial recognition systems, and
there is growing protest around the world.
Therefore:
1. We urge countries to suspend the further deployment of facial
recognition technology for mass surveillance;
2. We urge countries to review all facial recognition systems to
determine whether personal data was obtained lawfully and to
destroy data that was obtained unlawfully;
3. We urge countries to undertake research to assess bias, privacy
and data protection, risk, and cyber vulnerability, as well as
the ethical, legal, and social implications associated with the
deployment of facial recognition technologies; and
4. We urge countries to establish the legal rules, technical
standards, and ethical guidelines necessary to safeguard
fundamental rights and comply with legal obligations before
further deployment of this technology occurs.
https://thepublicvoice.org/ban-facial-recognition/
______
News Release, U.S. Travel Association
U.S. Travel Reacts to Suspension of Global Entry for New York Residents
WASHINGTON (February 6, 2020).--U.S. Travel Association Executive
Vice President for Public Affairs and Policy Tori Emerson Barnes issued
the following statement on the reported suspension of Global Entry and
several other trusted traveler programs for residents of the State of
New York:
``Travel should not be politicized. Trusted traveler programs enhance
our national security because they provide greater certainty regarding
a person's identity, citizenship, and criminal background. Suspending
enrollment in Global Entry and other trusted traveler programs only
undermines travel security and efficiency. We are in contact with the
Department of Homeland Security to convey this message.''
Contacts
Chris Kennedy: (O) 202.218.3603 (C) 202.465.6635
Tim Alford: (O) 202.218.3625 (C) 740.215.1290
###
U.S. Travel Association is the national, non-profit organization
representing all components of the travel industry that generates $2.5
trillion in economic output and supports 15.7 million jobs. U.S.
Travel's mission is to increase travel to and within the United States.
Visit www.ustravel.org.
Chairman Thompson. Mr. Wagner, if you are aware of any
notification requirements that a State would be noticed, I am
talking about the global entry situation because it looks like
New York is just the first of 1 or 2 others.
Since we have been sitting here, Mr. Cuccinelli has said
Washington State might be in a similar position.
I am just wanting to make sure that if this is the way
forward, then surely, in light of what Mr. Rose and some of the
other New Yorkers on this committee have said, there should be
some notice that this is about to happen and not just a press
conference.
So if you are aware of any, please get it back to us in the
committee. We would love to have it.
I thank the witnesses for their valuable testimony and the
Members for their questions.
The Members of the committee may have additional questions
for the witnesses, and we ask that you respond expeditiously,
in writing, to those questions.
Without objection, the committee record will be kept open
for 10 days.
Hearing no further business, the committee stands
adjourned.
[Whereupon, at 12:15 p.m., the committee was adjourned.]
[all]