Innocent man jailed for a week for face recognition system error

Face recognition systems using machine learning are used in systems such as automatic payment of subway fares and dining halls, as well as government surveillance networks and criminal investigations. On the other hand, the accuracy and privacy aspects of facial recognition systems are also being debated, with Italian regulatory authorities banning their use for purposes other than some criminal investigations and deterrence purposes. use is prohibited by It has been confirmed that misidentification is particularly likely to occur in the recognition of colored people, and a lawyer has announced a case in which an innocent black man was unfairly arrested due to an error in the face recognition system around the end of 2022.

JPSO used facial recognition to arrest a man.

Facial Recognition Error Sent Randall Reid to Jail, Lawyer Says

Black man wrongly jailed for a week after face recognition error, report says | Ars Technica

Randall Reed, who lives in Georgia, USA, was stopped by police while driving on a highway in Georgia in November 2022, and was arrested as one of the criminal groups involved in a series of thefts in Louisiana. . However, Mr. Reed was not involved in theft, and he never visited Louisiana. According to a Louisiana daily newspaper, the police issued a warrant for Mr. Reed as a suspect of theft by a facial recognition system, and arrested Mr. Reed based only on the facial recognition system. Police admitted their error and Reed was released about a week later.

Mr. Reed's lawyer, Tommy Calogero, said, 'Mr. Reed is 40 pounds (about 18 kilograms) lighter than the criminal who was shown on the surveillance camera, and the mole on Mr. Reed's face is the criminal in the image if humans compare it. The police acquiesced to this discrepancy, even though they knew he was a different person,' he said. In fact, numerous studies have shown that facial recognition systems successfully identify white male faces, but produce inaccurate results when identifying people of color and women. Although some law enforcement agencies acknowledge that facial recognition systems should only be used for clues, and should not be used as the sole basis for issuing arrest warrants, they have explicitly used facial recognition technology. There are not many cases where there are published rules to manage.

Claire Garvey, a training adviser to the National Association of Criminal Defense Lawyers, said, 'Failures in facial recognition systems continue to put innocent people in jail. 'The use of facial recognition is problematic, but it tends to grow.'

From 2020 to 2021, the police will use face recognition systems, as Congress across the United States will vote to ban face recognition in the police, and large cities in the United States will pass ordinances banning the use of face recognition systems one after another. There was a growing movement to ban its use. However, many states and cities such as California, Virginia, and New Orleans, Louisiana have withdrawn their bans on facial recognition systems, and the movement to limit facial recognition systems has lost momentum.

Large cities in the United States ban facial recognition systems one after another - GIGAZINE

by Mike MacKenzie

In New Orleans, which neighbors Metley, Louisiana, where Reed was accused of theft, Garvey said there are rules about how police use facial recognition systems, and facial recognition is used to get clues. It is police policy that only However, in Metairie, other parts of Louisiana, and much of the United States, law enforcement agencies have unrestricted use of facial recognition, and facial recognition is often involved in investigations. It is said that the prosecution does not even need to make public. Mr. Garvey argues that 'this violates the defendant's constitutional rights.'

Police in Baton Rouge, Louisiana, which issued a warrant for Reed's arrest, have been tight-lipped about the matter and have not detailed their policies regarding facial recognition.

in Software,   Security, Posted by log1e_dh