On a residential street in San Diego County, Calif., Chula Vista police had just arrested a young woman, still in her pajamas, for possession of narcotics. Before taking her away, Officer Rob Halverson paused in the front yard, held a Samsung Galaxy tablet up to the woman’s face and snapped a photo.
Halverson fiddled with the tablet with his index finger a few times, and – without needing to ask the woman’s name or check her identification – her mug shot from a previous arrest, address, criminal history and other personal information appeared on the screen.
Halverson had run the woman’s photograph through the Tactical Identification System, a new mobile facial recognition technology now in the hands of San Diego-area law enforcement. In an instant, the system matches images taken in the field with databases of about 348,000 San Diego County arrestees. The system itself has nearly 1.4 million booking photos because many people have multiple mug shots on record.
The little-known program could become the largest expansion of facial recognition technology by U.S. law enforcement. Amid an international debate over collecting and sharing huge amounts of data on the public, this pilot program is putting that metadata to use in the field in real time.
The use of this technology was rolled out without any public hearings or notice. In turn, the secrecy of the program has alarmed privacy experts and raised questions about whether San Diego is the leading edge of an alarming future – one in which few people escape cataloging in a government database.
Twenty-five local, state and federal law enforcement agencies – including U.S. Immigration and Customs Enforcement, the Border Patrol, the San Diego County Sheriff’s Department and San Diego State University – participate in the system. The project is coordinated by the San Diego Association of Governments and relies on a vast data-sharing program called the Automated Regional Justice Information System.
For some, the use of biometric technology by police represents a radical milestone in the militarization of American law enforcement.
For years, technology that was developed on the battlefield has been migrating into domestic police agencies. Since 9/11, America’s wars in Afghanistan and Iraq have sped up that transfer. Facial recognition technology, which has been widely used by the military, is the next frontier.
“What we’re seeing now is much more surveillance oriented, and it’s in the guise of preventative policing,” said Kevin Keenan, former executive director of the American Civil Liberties Union of San Diego & Imperial Counties. “It’s really this aspiration of prevention and social control through the monitoring of everyone’s every action and storage in perpetuity.”
San Diego’s program, if considered successful, easily could expand beyond the county’s borders.
The system’s mug shots are pulled from the statewide Cal-Photo law enforcement database, which also has access to 32 million driver’s license photos. And, according to a report by the Automated Regional Justice Information System, the county is looking at using mug shots from statewide gang and parolee databases, as well as information stored by the Department of Motor Vehicles.
Facial recognition graphic thumbClick for larger image
Click to view the full graphic.
The legality of law enforcement using facial recognition technology has not been tested in the courts. But a Privacy Impact Assessment, which the Automated Regional Justice Information System helped write, claims that photos of everyday people can be taken during “traditional police-civilian encounters.”
San Diego law enforcement agencies have used the facial recognition system since the beginning of this year, when 133 Galaxy tablets and smartphones were distributed to 25 law enforcement agencies around the region, according to documents obtained through a public records request by the Electronic Frontier Foundation, a San Francisco nonprofit that studies surveillance and privacy issues.
Compared with the number of arrests throughout the San Diego region, which has about 3.2 million residents, the system is rolling out with relatively modest numbers. In the first 10 months of 2013, officers ran 5,629 queries through the database.
The sheriff’s department and San Diego Police Department have the most devices, with 64 and 27 devices, respectively, and they have made nearly 2,000 queries into the system combined. The most active single user is an SDSU police officer who used a device 224 times from January to Oct. 30, according to the documents.
Officials with the sheriff’s department and San Diego Association of Governments declined requests for comment.
Law enforcement officials said the pilot program is a valuable tool to help them identify people who refuse to give their names or use fake identification. Immigration officials said they have used the system to help them when they encounter immigrants who don’t have authorization to be in the U.S.
“Photographs are neutral – you can’t say it’s racist when a camera is taking a neutral picture of someone,” said Halverson, the Chula Vista officer. “It’s hitting on certain points of contact. It’s doing a neutral analysis of a person.”
The software works by capturing a freeze frame of a live video feed, which then focuses on the face and uses the distance between the eyes as a baseline. An algorithm then analyzes unique textures and patterns on the face, cross-referencing the freeze frame at the rate of a million comparisons per second against the police mug-shot database that also has been processed by the software.
Halverson said he has used the system to identify injured people who were unresponsive and had no identifying documents. Other officers have been overwhelmingly positive, according to the Automated Regional Justice Information System.
One Immigration and Customs Enforcement agent who provided a testimonial said he used the device during a warrant sweep in Oceanside. While on the sweep, the agent wrote, his “ ‘spidy senses’ were tingling” about the immigration status of a neighbor of the person he was pursuing.
He decided to run the man’s picture through the facial recognition software. The agent discovered the man was in the country illegally and had a 2003 DUI conviction in San Diego.
“I whipped out the Droid (smartphone) and snapped a quick photo and submitted for search,” the immigration agent wrote in his testimonial for the Automated Regional Justice Information System. “The subject looked inquisitively at me not knowing the truth was only 8 seconds away. I received a match of 99.96 percent. This revealed several prior arrests and convictions and provided me an FBI #. When I showed him his booking photo, his jaw dropped.”
Law enforcement officials said the facial recognition software has built-in privacy safeguards. After an image taken in the field is run through the system, it is discarded by the central database, they said. They say it does not create a database of photos of people who are stopped by police and questioned.
“If you’re not in a criminal database, you have nothing to hide,” Halverson said.
However, during field tests with Chula Vista police, images taken by field officers were stored within individual tablets. It’s up to police to delete those photos on their own.
Officers who have used the system in San Diego rave about its precision in identifying people. But facial recognition technology remains imperfect. Documents obtained by the Electronic Privacy Information Center, a Washington nonprofit, show that the FBI’s facial recognition program could fail to identify the right person in 1 out of 5 encounters – potentially ensnaring innocent people in investigations.