Facial Recognition Technology – How Is It Being Used by Municipalities?
Beginning in early July 2019, San Francisco will start restricting its local police force’s access to surveillance equipment and will prohibit its use of facial recognition technology. This is the first of a growing number of municipalities to begin regulating or banning facial recognition technology by local police departments. In California, the cities of Berkeley and Oakland are currently debating the use of the software, and the state of Massachusetts is considering a complete state-wide ban. San Francisco chose to adopt the American Civil Liberties Union’s rationale for ceasing the use of the technology, citing its potential for abuse and inaccuracy and calling for other municipalities to do the same. Other states such as Pennsylvania, however, already use the technology and rely on it for anti-fraud protection and criminal law enforcement.
Facial recognition technology involves using biometrics, or various points on a person’s face, to create a data map of that person’s facial features from a picture or video. The software analyzes the geometry of someone’s face, including the distance between their eyes or from their hairline to their chin. Using these datapoints and measurements, the software compares the gathered facial data points with a known database of faces to find a match. According to a Georgetown University study, almost half of Americans have their face included on at least one searchable facial recognition database. This is perhaps unsurprising, given that as of 2018, 43 of the 50 states were using some kind of facial recognition technology. Further, the Federal Bureau of Investigations (FBI) maintains its own database of over 640 million faces.
Photos in the database can come from drivers’ licenses, mugshots, cell phone photos, or even photos posted on social media. In theory, anywhere your face has been recorded or photographed could be a potential spot for information-gathering. Certain states do limit the sources from which the photos in the database may originate. For instance, Maryland and Indiana limit their databases to drivers’ license records and Oregon limits its database to mugshots and photos obtained in criminal proceedings.
Facial recognition technology is often used as a tool to apprehend criminals who would otherwise remain unknown. This could mean matching identities to perpetrators of crimes such as identity theft or fraud, crimes caught on surveillance cameras at shops, or sexual assaults where the victim may have a photo of the assailant, but no full name. According to NBC News, the technology has allowed police to apprehend a “serial robber in Indiana, a rapist in Pennsylvania, a car thief in Maine, robbery suspects in South Carolina, a sock thief in New York City, and shoplifters in Washington County, Oregon.”
In Pennsylvania, PennDOT uses the technology to ensure that individuals do not fraudulently obtain more than one driver’s license. If the facial recognition program identifies what appear to be matching photos on a preexisting license and an applicant for a new or renewed license, PennDOT will issue an interim license for the applicant, and investigate further during a 15-day period. During that time, the matches are reviewed to determine if there is fraud or identity theft. If there is not, a permanent license is issued and there are no further problems.
Pennsylvania also uses a system called the Justice Network, or JNET, as a compendium of previously arrested or convicted criminals. JNET integrated its system with PennDOT’s previously existing repository in 2015 to expand its access to include over 39.5 million photos, according to JNET officials. Authorized law enforcement personnel can access the system for local or state use to compare photos from social media, texts, or surveillance footage to photos existing in the database to get leads on criminal suspects. Based on citizen privacy guidelines, officials can only use the technology to pursue a suspect during an investigation.
The technology is not perfect, however. Matches have not yet reached 100% accuracy—particularly for women and individuals of color. Further, if the photo or video being used for comparison is grainy, depicts a side-profile, or has obstructions covering the suspect’s face, like a mask or sunglasses, the results may be skewed. Even with a good photograph, investigators may not find promising matches until they reach the end of the results. Matches then require verification and further investigation to ensure the correct person is apprehended. The result of the facial recognition search itself does not give police probable cause to make an arrest.
Due to the controversial nature of the technology, many municipalities outside of Pennsylvania are regulating its use on their own, rather than waiting for the federal or state governments to act. Local ordinances and regulations can only do so much, however. Municipalities are without authority to limit the federal or state government’s use of facial recognition software, and they also cannot regulate commercial uses of the software for advertising purposes. They can regulate use of the software by the local municipal police force. Municipalities considering regulating or banning the use of facial recognition software should weigh the benefits against privacy concerns associated with use of the technology in making their determinations.
With contribution from Sarah Rothermel, J.D. Widener Law Commonwealth.