US law enforcement dispenses ammo to facial recognition critics

America’s embrace of surveillance technology and a regular series of arrests based on false matches are keeping the question of whether U.S. police use of facial recognition is appropriate on the front burner.

Biometrics use is increasing amongst police around the world, according to the latest industry survey from the Biometrics Institute. Public education is needed for greater acceptance of biometrics, 84 percent of survey respondents say. But other stakeholders clearly need more education too.

The latest evidence is a parade of incompetence and ignorance about biometric technology featuring bounty hunters, police, attorneys and local government employees.

Texas bounty hunters

Given the trouble many U.S. police have had following their force’s policies around facial recognition use, it can be little surprise that a trio of bounty hunters in Texas are have been charged with felonies after misusing the technology.

The three people in Richardson, Texas kidnapped a local man in the belief that he was a murder suspect. He was not. The would-be vigilantes used unnamed facial recognition software, according to KRLD 1080, to match a photo of the suspect. The search returned a potential match but with a confidence score of only 79 percent, which was reported by the local news outlet as “accuracy.”

Even back in 2020, Amazon was recommending a 95 percent confidence threshold for law enforcement uses of its Rekognition facial recognition software. Indian police were  criticized for using 80 percent as a minimum confidence level in 2022.

The trio detained the individual after using stun grenades in the man’s garage and pointing assault rifles at him, and called police. A police officer fingerprinted him with a mobile scanner, and cleared him based on more reliable biometric practice.

The bounty hunters have been charged with aggravated kidnapping.

NYPD

A man arrested by a New York Police Department detective for allegedly exposing himself to a woman was first identified with a false positive facial recognition match, The New York Times reports.

A difference in stature of 8 inches of height and 70 pounds would be enough for most casual observers to tell Trevis Williams from the suspect, but not the NYPD.

Williams had been previously arrested for a case that was subsequently dismissed.

The department’s policy for facial recognition use appears to have been followed, with a human reviewer selecting Williams for placement in a photo lineup, and the victim identifying Williams from among six photos. That identification provided the probable cause for arrest.

Unlike in New York, police in Indiana and Detroit are required to have additional evidence to support a facial recognition match before using it in a suspect lineup. The requirement in Detroit is part of a legal change in Michigan prompted by the wrongful arrest of Robert Williams.

Beyond the facial recognition element of the investigation, however, more questions arise. The suspect was an Amazon delivery driver, but police did not contact the company to find out inquire about the company’s employee.

NYPD Spokesman Brad Weekes disputed that NYPD has made false arrests based on facial recognition matches. Public defenders group Legal Aid says it knows of a 2022 incident in which a man was identified with facial recognition and detained for a month before his case was dismissed, and warns there may be more, yet unknown misidentifications.

Of the 10 false arrests following facial recognition matches publicized so far in the United States, all but one were of Black people.

Public defender

A public defender in Chicago has requested that all evidence stemming from a facial recognition match of her client, who allegedly robbed and murdered an exchange student, be thrown out.

The Chicago Sun-Times reports that Quandee Semrow is arguing that facial recognition is unreliable and its use may have tainted evidence. The attorney is asking for the judge to either force disclosure from the prosecution about how the technology was used, or block evidence derived from it.

Cook County Circuit Judge John F. Lyke noted that the motion is likely moot, given the prosecution’s abundant evidence.

Across the state border in Wisconsin, Milwaukee police are still considering expanding their facial recognition capabilities, in the face of growing opposition that now includes the Milwaukee Equal Rights Commission and the Milwaukee Common Council, according to Wisconsin Watch.

Municipal employees

Town employees in Houlton, Maine have been caught searching municipal security cameras thousands of times to find the location of people or vehicles over the past year, The County reports.

Staff from the public works, recreation and civic center departments joined others from the police and code enforcement in using cameras or searching their footage during and beyond working hours. Officials turned on the cameras to watch a live video feed in real time on 56,000 occasions. The numbers would presumably be significantly higher if the cameras hadn’t been shut down for a legal review from February to July of this year.

Unidentified users also accessed the system at least twice, after non-police employees gave out login credentials.

The Verkada cameras come with facial recognition capabilities, but The County says it is unclear whether the feature is in use. If so, it could violate Maine law. Town Council also passed a policy barring the use of facial recognition.

Article Topics

biometric identification  |  biometric matching  |  biometrics  |  facial recognition  |  false arrest  |  law enforcement  |  NYPD  |  police  |  United States