AI/Machine Learning, Cybersecurity

The GAO Report’s Insight on the Expansion of Facial Recognition Tech


Clearview AI, a third-party facial recognition contractor for the government, claims that nearly everyone in the world will be identifiable upon achieving the collection of 100 billion photos for their facial recognition system by the beginning of 2023.[1] Statements like this and other reports have generated increased concern and investigations to determine the extent of facial recognition technology and what role it should play within the government.

The GAO Report and Defining FRT

The most recent Government Accountability Office (GAO) report published in August 2021 followed heightened public concern regarding the use of facial recognition technology (FRT) by federal government agencies. At the request of Congress, the report surveyed 24 agencies listed in the Chief Financial Officers Act of 1990 over the 2020 fiscal year.[2]

Facial recognition, one of several types of biometric identifiers like DNA and fingerprints, maps out a mathematical formula of an individual’s face. This faceprint or facial signature creates a unique way to verify identity. A FRT system’s algorithm then compares the information to find matches within the database. Experts attribute some of the controversies surrounding FRT to a misunderstanding of its definition and the differentiations between functions. Facial verification seeks to confirm identity by comparing two photos to ensure they are the same person. Facial identification pulls from a database for the closest matches to a photo. Facial analysis classifies a face’s personal characteristics and determines expressions.[2] The report distinguishes between the different facets of the technology but for the purpose of the study, the GAO includes all of the above under the umbrella of FRT.

Federal Usage of FRT

In the GAO survey, researchers evaluated current usage and discovered that of the 24 agencies surveyed, 18 used facial recognition technology during 2020 for one or more purposes.[3]

Agencies that applied FRT include:

  • Department of Agriculture
  • Department of Commerce
  • Department of Defense
  • Department of Energy
  • Department of Health and Human Services
  • Department of Homeland Security
  • Department of the Interior
  • Department of Justice
  • Department of State
  • Department of the Treasury
  • Department of Veterans Affairs
  • U.S. Agency of Internal Development
  • Environmental Protection Agency
  • General Services Administration
  • National Aeronautics and Space Administration
  • National Science Foundation
  • Office of Personnel Management
  • Social Security Administration

Carahsoft GAO Report Facial Recognition Tech Blog Embedded Image 2022 The top three uses include digital access or cybersecurity, domestic law enforcement, and physical security. Sixteen agencies attested to incorporating FRT for digital access or cybersecurity which mostly comprised employees unlocking agency phones. FRT was used by six agencies to generate leads in criminal investigations by identifying victims or criminals and creating possible person of interest lists. Five agencies employed FRT for additional physical security such as building access. Other uses included border and transportation security, national security and defense, research and development, and medical assessment.[3]

To run these recognition searches, the government continues to build internal databases, purchase external commercial software, and collaborate with other internal agencies to cross check and obtain additional records. The Defense, Justice, and Homeland Security departments own two thirds of federal FRT systems.[4]

The most prominent third-party system government agencies use, Clearview AI, is wrapped in lawsuits and controversy over their collection of photos pulled from social media to build their database without users’ knowledge or consent. Some statements claim up to 3,100 or 17% of federal, state, county, and municipal agencies employed Clearview’s technology.[5] Despite public criticism, more government agencies hope to use Clearview’s system to draw from their exponentially expanding 2022 repository of over 10.5 billion faces.[6]

Government Expansion of FRT

According to the GAO report, the Agriculture, Commerce, Defense, Homeland Security, Health and Human Services, Interior, Justice, State, Treasury and Veterans Affairs departments intend to continue expanding facial recognition usage. By 2023, these government agencies plan to have developed or purchased 17 additional FRT systems, accessed two more local systems, and begun two more Clearview AI contracts.[2]

Over the 20 years that government agencies and police have utilized FRT, they were generally restricted to only scanning government-provided images like mug shots and driver’s license photos.[7] Now, access to systems like Clearview have changed that. Most of the expansion plans fall under three categories: system updates, problem solving, and new features.

The State Department wants to implement a new FRT-based program using the Personal Identification Secure Comparison and Evaluation System (PISCES) border management system to evaluate travelers for suspected terrorists. The DHS plans to pilot a similar program with TSA and forecasts by 2023, facial recognition will be used on 97% of travelers.[8] Other additions include the U.S. Marshals Service plans to construct a touchless prisoner identification system for the jail and prison networks. Pending funding approval, the Department of Agriculture hopes to use FRT to scan live surveillance videos for people on watchlists.

Ten agencies working on research and development seek to improve systems’ accuracy, reduce false match rates by increasing the algorithms’ training database, and teach FRT to recognize individuals wearing masks.[5Through their research, the State Department hopes to eliminate the challenge of aging confusing recognition systems. The Department of Transportation’s research would allow for FRT to evaluate commercial truck drivers, train drivers, and air traffic controllers’ eyes for signs of distraction, drowsiness or fatigue.[9] While government agencies have extensive plans for FRT, the path forward with expansion abounds with many challenges.

Challenges Confronting Future Use of FRT

The major concerns with government FRT and protests to its expansion stem from issues with racial bias, inaccuracy, loss of privacy, and a lack of oversight. Initial reviews into FRT systems in the late 2010s returned some troubling data regarding the technology’s accuracy and bias. Recent studies reveal significant system improvement; however, other challenges remain at the forefront.

Results of the GAO survey showed that 13 agencies were unaware of their employees’ usage of non-federal facial recognitions systems. This discovery prompted the GAO’s recommendation that agencies initiate a tracking mechanism to close this oversight gap and address the risks involved with employing non-federal systems like Clearview.[10] In early 2021 before the GAO report was released, over 35 civil rights groups had already petitioned for legislation that would halt government usage of FRT and finalize legal standards of usage.[10] Currently, no federal regulations are in place to provide accountability regarding these issues and local bans that many cities and states have resorted to do not inhibit federal usage. In the absence of congressional laws, Amazon and Microsoft enforced moratoriums and temporary bans on selling their facial recognition technology to the government.[11]

Uncertainty shrouds the future of FRT, but until official regulations fall into place, government agencies are pursuing the latest system updates and upgrades. The GAO report offers a better frame of reference on current usage and insight into government agencies’ plans for growth while legislation is pending.

 

Visit our website to learn how Carahsoft is helping government agencies connect technology and industry partners with best-of-breed artificial intelligence, machine learning, and high performance computing capabilities to meet mission needs.

 

[1] “Clearview AI plans to put almost every human face in its database,” Silicon Republic, https://www.siliconrepublic.com/enterprise/clearview-ai-100-billion-photos-facial-recognition-database

[2] “FACIAL RECOGNITION TECHNOLOGY: Current and Planned Uses by Federal Agencies,” United States Government Accountability Office, https://www.gao.gov/assets/gao-21-526.pdf

[3] “Facial Recognition Technology: Current and Planned Uses by Federal Agencies,” United States Government Accountability Office, https://www.gao.gov/products/gao-21-526

[4] “Summary of the GAO Report on Federal Use of Facial Recognition Technology,” Lawfare, https://www.lawfareblog.com/summary-gao-report-federal-use-facial-recognition-technology

[5] “US government agencies plan to increase their use of facial recognition technology,” Technology Review, https://www.technologyreview.com/2021/08/24/1032967/us-government-agencies-plan-to-increase-their-use-of-facial-recognition-technology/

[6] “Clearview AI aims to put almost every human in facial recognition database,” Ars Technica, https://arstechnica.com/tech-policy/2022/02/clearview-ai-aims-to-put-almost-every-human-in-facial-recognition-database/

[7] “The Secretive Company That Might End Privacy as We Know It,” The New York Times, https://www.nytimes.com/2020/01/18/technology/clearview-privacy-facial-recognition.html

[8] “What is Facial Recognition – Definition and Explanation,” Kaspersky, https://www.kaspersky.com/resource-center/definitions/what-is-facial-recognition

[9] “Federal government to expand use of facial recognition despite growing concerns,” The Washington Post, https://www.washingtonpost.com/technology/2021/08/25/federal-facial-recognition-expansion/

[10] “Calls for Biden to ban facial recognition grow after GAO report’s findings,” Daily Dot, https://www.dailydot.com/debug/facial-recognition-gao-biden/

[11] “Rules around facial recognition and policing remain blurry,” CNBC, https://www.cnbc.com/2021/06/12/a-year-later-tech-companies-calls-to-regulate-facial-recognition-met-with-little-progress.html

Related Articles