Spotlight on Surveillance

November 2005:
Facial Recognition Systems Have an Ugly Effect on Personal Privacy

 

The federal government is spending an increasing amount of money on surveillance technology and programs at the expense of other projects. EPIC's "Spotlight on Surveillance" project scrutinizes these surveillance programs. For more information, see previous Spotlights on Surveillance.¹

This month, Spotlight focuses upon the government’s use of facial recognition systems.² The Departments of State, Energy, Justice and Defense have spent at least $47 million on such systems.³ In Fiscal Year 2006, the federal government plans to add facial recognition checks to all visa applications, which already include fingerprint biometrics.⁴ This is despite the Government Accountability Office’s estimate that incorporating biometric systems into visas would cost from $1.3 billion to $2.9 billion for startup, and $700 million to $1.5 billion for annual operating costs.⁵ Federal funds have been used by cities and states to buy facial recognition devices for motor vehicle and police departments.⁶ New U.S. passports and national identification cards created under the REAL ID Act of 2005 will both include digital photographs that can be linked to facial recognition systems.⁷ However, several tests, including those conducted by the National Institute for Standards and Technology (NIST) and the Department of Defense, show that facial recognition systems can be easily befuddled by uncooperative subjects and changes in the environment, such as positioning or lighting.⁸ Such facial recognition systems create significant privacy risks because the technique is surreptitious, the prospects for extensive profiling are clear, and there are no laws that currently regulate these systems to prevent abuse.

Face recognition technology vendor Identix admits that when there is an angle difference between images "[f]rom 15 to 35 degrees, the face recognition discrimination power decreases," and "[a]ngled faces beyond 35 degrees do not match well to frontal faces with" its current technology. See Identix's FAQs about facial recognition (pdf). Source: http://www.identix.com/

In simplified terms, a facial recognition system compares a digital pattern (a computerized estimate) of the face of a person walking down the street or standing in a crowd to images stored in databases. Changes in the environment, such as positioning, lighting and shadows, can affect biometric data collection.⁹ Changes in facial expression, angle of head to camera, hairstyle, and beards are particularly effective at confounding facial recognition systems.¹⁰ In one test, the Government Accountability Office found that a person “moving his head slightly left and right also fooled the system.”¹¹ The Defense Department’s face recognition technology tests found that a 45-degree difference in position between the current and database images renders the technology ineffective, and a 15-degree difference is all that is needed to adversely affect the facial recognition technology.¹²

In Congressional testimony, EPIC Executive Director Marc Rotenberg said the effectiveness of a system of biometric identification would be determined by how the system is set up, protected and maintained.¹³ He explained there are several ways to compromise the effectiveness of a biometric system – by false identification at enrollment, physical alteration of a personal biometric, stewing the sample collection by not cooperating, and hacking into of falsifying data.¹⁴ Face recognition system errors lead to innocent people being falsely matched to watchlists or databases, and suspects being able to pass through the system unrecognized.

In 2002, NIST conducted a comprehensive study of facial recognition systems. NIST found that the recognition rate for faces captured in uncontrolled environments, such as outdoors, could be as low as 50%.¹⁵ NIST also determined that, because of the high failure rates when applied to large groups of people, facial recognition was not a viable technology for “large-scale identification.”¹⁶ Time also affects facial recognition systems. The longer the time between the original photograph in the database and the new image captured, the less likely the facial recognition system will make a correct match.¹⁷

According to biometrics technology company Viisage, its FaceFINDER system "scans crowds of people and matches individuals to selected faces previously stored in an image database," and "[w]ithout the visitor specifically being requested to look into the camera, the system can be deployed effectively." [click to view full brochure in pdf] Source: http://viisage.com

Tampa is one of the U.S. cities that has used facial recognition technology in concert with camera surveillance systems to surreptitiously scan the public.¹⁸ Tampa used a facial recognition camera system to scan the faces of people at the Super Bowl in 2001, and the police department subsequently used such a system to watch people in the Ybor City nightlife district.¹⁹ A report by the American Civil Liberties Union found that:

[t]he system has never correctly identified a single face in its database of suspects, let alone resulted in any arrests … [and its] photographic database contains a broader selection of the population that just criminals wanted by the police, including such people as those who might have ‘valuable intelligence’ for the police or who have criminal records.²⁰

In August 2003, Tampa stopped using the system, supplied by Identix, because of its failures. “It’s just proven not to have any benefit to us,” said a police department spokesman.²¹

The mission creep of the Tampa system is an example of the privacy risks created by such facial recognition systems. What began as system to catch criminals became a system to find people who might have information for which the police are searching. That is a poor reason to invade the privacy of the general public. With such systems, a person can be scanned without her knowledge or consent. A person’s “suspect activity” may be no more than walking around a popular nightlife area.

Boston’s Logan Airport ran two separate tests of face recognition systems – supplied by Viisage Technology and Identix – in 2002.²² Both systems had such high error rates that the airport dropped the technology.²³ Facial recognition systems also failed tests at airports in Dallas/Fort Worth, Fresno, Calif., and Palm Beach County, Fla.²⁴ One glaring example of a facial recognition system failure occurred when two people swapped passports at an Australia airport as a joke. Their deception was not caught by the facial recognition systems.²⁵ Yet Australian authorities continue to test the technology, and the “trial” has encompassed three years.²⁶

Despite this poor history, facial recognition systems still are being used. The Defense Department spent $1.6 million to pay Identix for researching facial recognition technology.²⁷ The Massachusetts Registry of Motor Vehicles recently spent $1.5 million in federal funds for facial recognition systems supplied by Digimarc.²⁸ Virginia Beach, Va. used $150,000 in federal funds, plus $50,000 of its own funds, for a facial recognition technology-enabled camera surveillance system supplied by Identix.²⁹ The Texas Department of Public Safety recently contracted with Identix to pay $1.8 million to upgrade its systems to include facial recognition capabilities.³⁰ Pinellas County, Fla. used $8 million in federal grants to outfit patrol cars, jails and an airport with computerized facial recognition systems supplied by Viisage Technology.³¹ The Los Angeles police department is using handheld facial recognition devices supplied Neven Visions.³² Once again, a person could be merely walking down the street and have a facial recognition device focused on her.

A Department of Defense study shows that changes in lighting and time lapse between the capture of each image can significantly affect face recognition performance. Source: P. Jonathon Phillips, et al., An Introduction to Evaluating Biometric Systems, Computer at 61 (Feb. 2000) available at http://www.frvt.org/DLs/FERET7.pdf and http://www.epic.org/privacy/surveillance/spotlight/1105/feret7.pdf

The two major companies supplying face recognition technology are Massachusetts-based Viisage Technology and Minnesota-based Identix. Identix’s technology includes: FaceIt SDK, used in developing enrollment, one-to-one and one-to-many applications; FaceIt Surveillance SDK, when added on to FaceIt SDK, used to develop facial screening applications; FaceIt Quality Assessment SDK; ABIS System, a web-based multi-biometric search engine used to conduct one-to-many searches; FaceIt ARGUS, facial screening software for real-time verification.³³ Identix’s facial recognition systems have not solved the technology’s failures. The company admits that when there is an angle difference “[f]rom 15 to 35 degrees, the face recognitions discrimination power decreases,” and “[a]ngled faces beyond 35 degrees do not match well to frontal faces with” its current technology.³⁴

Viisage’s facial recognition systems include: FaceEXPLORER for large digital image database management, FaceTools SDK designed for non-distributed environments, FaceFINDER for screening and surveillance, and FacePASS for access control.³⁵ Viisage’s FaceFINDER system is especially troubling. The facial recognition system “scans crowds of people and matches individuals to selected faces previously stored in an image database,” and “[w]ithout the visitor specifically being requested to look into the camera, the system can be deployed effectively.”³⁶

	Places in the U.S. where face recognition has been, is being or will be used: Federal government: Departments of Defense, Energy, Homeland Security, Justice, State California: Fresno; Los Angeles Florida: Palm Beach County; Pinellas County; Tampa Massachusetts: Boston; state motor vehicle registry Texas: Dallas/Fort Worth Virginia: Virginia Beach
Source: News reports

Spotlight on Surveillance has previously discussed the increasing use of camera surveillance systems throughout the United States.³⁷ Cities such as Chicago and Baltimore used money from homeland security grants to create networks of surveillance cameras to watch over the public in the streets, shopping centers, at airports and more.³⁸ However, studies have found that such surveillance systems have little effect on crime, and that it is more effective to place more officers on the streets and improve lighting in high-crime areas.³⁹

Facial recognition software is also increasingly used internationally as well. The International Civil Aviation Organization has set standards suggesting fingerprints and face recognition technology be incorporated into all passports.⁴⁰ The United Kingdom, which has an extensive camera surveillance network, has added face recognition technology to the system.⁴¹ London has 200,000 cameras, and more than 4 million cameras have been deployed throughout the country.⁴² It is estimated that there is one camera for every 14 people, and the average Briton is seen by 300 cameras per day, according to estimates.⁴³

The State Department has recently issued new standards for U.S. passports that mandate digital photographs. It could add these photographs to its current large database of digital images, collected during the past 10 years from passport applications submitted by 70 million Americans.⁴⁴ The recently passed REAL ID Act mandates digital photographs in its national identification cards. Soon, every person who holds a U.S. passport or a REAL ID card will have his or her face placed into a database. Internationally, “[t]here are 1.2 billion digitized photos of people in databases around the world,” according to the president of Identix.⁴⁵

Facial recognition systems have already been used in one test to scan crowds for more than just suspected criminals or terrorists, but also for “such people as those who might have ‘valuable intelligence’ for the police,” whatever “valuable intelligence” may mean. It is a small step for the government to take to continually use the networks of facial recognition-enabled surveillance cameras to surreptitiously search crowds for anyone in a database containing millions of people’s faces, though they are suspected of no crime and no legal basis to monitor them has been established.

EPIC recommends that funding for these systems be suspended until a comprehensive evaluation of their utility and effectiveness is completed. EPIC further recommends that a Privacy Impact Assessment that specifically examines the likelihood of “mission creep” be conducted before any facial recognition system is deployed. Finally, EPIC recommends that legal guidelines are established for the use of facial recognition systems.

---

¹ EPIC, “Spotlight on Surveillance,” available at http://www.epic.org/privacy/surveillance/spotlight/

² See, EPIC and Privacy International, Privacy and Human Rights: An International Survey of Privacy Laws and Developments at 103-104 (EPIC 2004) available at http://www.privacyinternational.org/phr/; EPIC’s Facial Recognition page at http://www.epic.org/privacy/facerecognition/.

³ Letter from the General Accounting Office to Dick Armey, Majority Leader, House of Representatives at 4 (Mar. 14, 2002).

⁴ Department of State and Agency for International Development, Performance Summary: Fiscal Year 2006, 11206 (Feb. 7, 2005) available at http://www.state.gov/s/d/rm/rls/perfplan/2006/pdf/ and http://www.epic.org/privacy/surveillance/spotlight/1105/statefy06.pdf; Department of State, Daniel B. Smith, Principal Deputy Assistant Secretary for Consular Affairs, Testimony on the Department of State’s Border Security Program: Remarks before the House Committee on International Relations Subcommittee on Africa, Global Human Rights and International Operations (May 12, 2005) available at http://www.state.gov/r/pa/ei/othertstmy/48735.htm; for more on border security, see EPIC’s US-VISIT page at http://www.epic.org/privacy/us-visit/.

⁵ The Government Accountability Office was previously called the General Accounting Office. General Accounting Office, Technology Assessment: Using Biometrics for Border Security, GAO 03-174 (Nov. 2002) available at http://www.gao.gov/cgi-bin/getrpt?GAO-03-174 and http://www.epic.org/privacy/surveillance/spotlight/1105/gao03174.pdf (hereinafter “GAO Biometrics Report”).

⁶ Infra.

⁷ Public Notice of Final Rule: Electronic Passport, 70 Fed. Reg. 61553 (Oct. 25, 2005) available at http://edocket.access.gpo.gov/2005/05-21284.htm; REAL ID Act of 2005, Pub. L. No. 109-13 (2005) available at http://thomas.loc.gov/cgi-bin/bdquery/z?d109:h.r.00418:; for more on these plans, see EPIC’s RFID and National ID pages at http://www.epic.org/privacy/rfid/ and http://www.epic.org/privacy/id_cards/.

⁸ P. Jonathon Phillips, et al., National Institute of Standards and Technology, Face Recognition Vendor Test 2002: Overview and Summary (Mar. 2003) available at http://www.frvt.org/DLs/FRVT_2002_Overview_and_Summary.pdf and http://www.epic.org/privacy/surveillance/spotlight/1105/nist0303.pdf (hereinafter “NIST Face Recognition Study”).

⁹ GAO Biometrics Report at 173, supra note 4.

¹⁰ Id.

¹¹ Id.

¹² P. Jonathon Phillips, et al., An Introduction to Evaluating Biometric Systems, Computer at 61 (Feb. 2000) available at http://www.frvt.org/DLs/FERET7.pdf and http://www.epic.org/privacy/surveillance/spotlight/1105/feret7.pdf.

¹³ Statement of Marc Rotenberg, Executive Director, Electronic Privacy Information Center, and Carla Meninsky, EPIC IPIOP Fellow, at a Joint Hearing on Identity Theft Involving Elderly Victims Before the Special Committee on Aging (July 18, 2002) available at http://www.epic.org/privacy/biometrics/testimony_071802.html.

¹⁴ Id.

¹⁵ NIST Face Recognition Study at 2, supra note 7.

¹⁶ National Institute of Standards and Technology, Summary of NIST Standards for Biometric Accuracy, Tamper Resistance, and Interoperability, at 6 (Nov. 13, 2002) available at ftp://sequoyah.nist.gov/pub/nist_internal_reports/NISTAPP_Nov02.pdf and http://www.epic.org/privacy/surveillance/spotlight/1105/nist1102.pdf

¹⁷ NIST Face Recognition Study at 2, supra note 7.

¹⁸ Thomas W. Krause, City Unplugs Camera Software, Tampa Tribune, Aug. 20, 2003.

¹⁹ Id.

²⁰ Jay Stanley and Barry Steinhardt, American Civil Liberties Union, Drawing a Blank: The failure of facial recognition technology in Tampa, Florida at 1 (Jan. 3, 2002) available at http://www.epic.org/privacy/surveillance/spotlight/1105/aclu0302.pdf (originally available at http://archive.aclu.org/issues/privacy/drawing_blank.pdf).

²¹ Thomas W. Krause, City Unplugs Camera Software, Tampa Tribune, Aug. 20, 2003.

²² Richard Willing, Airport anti-terror systems flub tests, USA Today, Sept. 2, 2003.

²³ Id.

²⁴ Id.

²⁵ Karen Dearne, SmartGate joke a serious concern, The Australian, Mar. 4, 2003.

²⁶ Fed: Govt extends trial of faster passenger screening, Australian Associated Press, Feb. 2, 2005; Australia steps up border security with biometrics trial, Kyodo News Service, Sept. 29, 2005.

²⁷ Identix Gets DOD $1.6 Million For Researching Facial Recognition Technology, FinancialWire, Sept. 8, 2005.

²⁸ Jesse Noyes, Registry eyes new line: Is that the real you?, Boston Herald, Oct. 19, 2005.

²⁹ Fred Guterl and William Underhill, Taking a Closer Look, Newsweek, Mar. 8, 2004.

³⁰ Identix Recognition Technology Selected For $1.8 Million Upgrade By Texas Department, Financial Wire, Sept. 15, 2005.

³¹ Richard Winton, ID System Gets in Face of Criminals, L.A. Times, Dec. 25, 2004.

³² Id.

³³ Identix Inc.: http://www.identix.com.

³⁴ Identix Inc., FaceIt® G6 Frequently Asked Technical Questions (last updated May 20, 2005) available at http://www.identix.com/trends/faqs/faceit_faqs.pdf and http://www.epic.org/privacy/surveillance/spotlight/1105/facefaqs.pdf.

³⁵ Viisage Technology: http://www.viisage.com.

³⁶ Viisage Technology, FaceFinder 2.5 Data Sheet (2004) available at http://www.viisage.com/en/data/pdf/en_datasheet_facefinder_2-5.pdf and http://www.epic.org/privacy/surveillance/spotlight/1105/facefinder.pdf.

³⁷ EPIC, Spotlight on Surveillance: More Cities Deploy Camera Surveillance Systems with Federal Grant Money (May 2005) at http://www.epic.org/privacy/surveillance/spotlight/0505/.

³⁸ Privacy and Human Rights at 102-103, supra note 1.

³⁹ See generally Privacy and Human Rights at 95-104,supra note 1; Brandon C. Welsh and David P. Farrington, Home Office Research, Development and Statistics Directorate, Crime prevention effects of closed circuit television: a systematic review, Research Study 252 (Aug. 2002) available at http://www.homeoffice.gov.uk/rds/pdfs2/hors252.pdf; National Association for the Criminal Rehabilitation of Offenders, To CCTV or not to CCTV? A review of current research into the effectiveness of CCTV systems in reducing crime (June 28, 2002) available at http://www.epic.org/privacy/surveillance/spotlight/0505/nacro02.pdf.

⁴⁰ International Civil Aviation Organization, Biometrics Deployment of Machine Readable Travel Documents: Technical Report Version 2.0, Doc. 9303 (2004) available at http://www.icao.int/mrtd/publications/doc.cfm.

⁴¹ Privacy and Human Rights at 703-704, supra note 1.

⁴² Fran Spielman and Frank Main, City plans camera surveillance web, Chicago Sun-Times, Sept. 10, 2004; see generally Privacy International, Overview: CCTV and Beyond available at http://www.privacyinternational.org/article.shtml?cmd[347]=x-347-65433.

⁴³ Id.; Tara Burghart, Chicago Mayor Unveils Surveillance Plan, Associated Press, Sept. 10, 2004.

⁴⁴ Eric Lipton, Hurdles for Technology in U.S. Security Efforts, N.Y. Times, Aug. 10, 2005.

⁴⁵ Barnaby J. Feder, Technology Strains to Find Menace in the Crowd, N.Y. Times, May 31, 2004.