Facial recognition pilot to identify New York drivers reports zero percent success rate
The first phase in a proof of concept test of biometric facial recognition at the Robert F. Kennedy Bridge by New York’s Metropolitan Transportation Authority have failed to detect a single face, the Wall Street Journal reports.
In an email seen by the Journal, an MTA official wrote to a senior official in the administration of Governor Mario Cuomo that the initial testing period was “completed and failed with no faces (0%) being detected within acceptable parameters.” A second hard drive of images has been sent for analysis and the MTA is procuring additional cameras to expand the program, according to the email.
“We are testing the technology, and all others that will help us keep New Yorkers safe, while protecting their civil liberties,” a spokesperson for the governor told the Journal.
New York Civil Liberties Union technologist Daniel Schwarz notes that there is a significant difference between using facial recognition for investigations and “pervasive, real-time surveillance of everyone passing by.”
An MTA spokesperson said the pilot at RFK and other bridges and tunnels is ongoing, as is the evaluation of the technology. Former MTA Chairman Joe Lhota said the system would be used to crack down on people evading fares by covering up their license plates, but spokesperson Maxwell Young says that it is only being used for security. Lhota says the agency takes civil liberties seriously, does not share data with law enforcement or anyone outside of the pilot, and that it can only be accessed by a “small handful” or employees. He also said that the MTA has paid IDEMIA about $25,000 for the pilot. The company did not respond to the Journal’s request for comment.
The pilot was launched in September, and faced criticism from the NYCLU and State Assembly.
Facial recognition researcher Hector Santos-Villalobos of Oak Ridge National Laboratory in Tennessee says a recent study at the laboratory achieved identification accuracy of higher than 80 percent for images captured through the windshields of slow-moving vehicles. Santos-Villalobos says that detecting faces in vehicles moving at higher speeds is more challenging, though possible.