5 things to know – San Bernardino Sun


5 things to know – San Bernardino Sun

The Chino Police Department started using facial recognition software in a trial run in 2019, and it plans to ask the City Council to approve it for full-time use.

Chino Police Chief Wes Simmons and Sgt. Dustin Tomicic recently held an event to show the public how the software functions, and answered questions about it.

“We want to take away some of the myths about the system and share how we’d use it within the Chino Police Department,” Simmons said.

1. How does facial recognition software work?

The software that the Chino Police Department used in its trial run homes in on 106 parts of the face from a photo and tries to find similarities between that photo and photos taken of suspects booked into jail or prison. It can establish a match in roughly three to five seconds.

The technology used in the trial run was supplied by AFR Engine, a firm based in Rancho Cucamonga specializing in facial recognition software.

Chino police only look to match against booking photos from California law enforcement agencies, Tomicic said. The software would not be used to identify witnesses, only suspects, according to police.

“The only caveat to that is for people who are unable to identify themselves that we’re looking to identify,” Tomicic said. “Such as deceased or mentally incapacitated people.”

Simmons and Tomicic said the software would only be used on photos or still images from video that had been voluntarily submitted, and that it wouldn’t be matched against photos from the Department of Motor Vehicles or social media.

Photos or still images from videos can come from cell phone cameras, business surveillance camera or even doorbell cameras.

However, if photos or videos are not voluntarily provided to police, it is possible that they could be obtained via a warrant.

“It’s not something we regularly do,” Tomicic said. “We seek the assistance of our community and try to get them to provide the information we need to follow up on these leads.”

“The courts provide the warrant process,” Simmons said. “I don’t know of any case where we’ve done that. Here’s the deal, if someone doesn’t want to give us photo or video, they’ve probably already long deleted it.”

2. What happened during the trial run?

Tomicic said that 127 images were submitted to police during the trial run. Of those 127 images, 98 resulted in no matches. Of the remaining 29, 20 matched with at least one person and a suspect was identified after additional investigation. Those 20 cases included robberies, burglaries, a shooting, grand thefts, assault with a deadly weapon, grand theft auto, felony evading and fraud.

“Nine more suspects were identified as a potential match, but the investigation is ongoing,” he said.

3. Facial recognition software is already prevalent in law enforcement

The technology is in use around Southern California and the U.S. Tomicic said the San Bernardino County Sheriff’s Department has been using it for at least 15 years, and the Los Angeles County Sheriff’s Department has been using it since at least 2009. Officials from both the Pomona and Riverside police departments said those agencies use facial recognition software from county sheriff’s departments.

However, California Assembly Bill 1215 prohibits the use of facial recognition technology on body-worn cameras. AB 1215, The Body Camera Accountability Act, was signed by Gov. Gavin Newsom on Oct. 8, 2019. The bill is set to expire on Jan 1, 2023.

4. The software is controversial

A June 2019 survey by the Pew Research Center showed a majority of Americans – 56% – trust law enforcement agencies to use facial recognition software in a responsible way.

But Mohammad Tasjar, a staff attorney for the ACLU in Southern California, said the technology can lead to inaccuracies and the possibility for misuse.

“There are concerns about accuracy and concerns about privacy and civil rights,” Tasjar said. “The way that you turn a face into code is one that requires training the system. You train the system by looking at a lot of faces. If you train your system on data sets that are biased, you’re likely to get outcomes where the computer thinks one person is somebody else. Math is not a panacea.”

“And even if everything was accurate, it doesn’t address the principal problem. It’s too invasive to be put in the hands of law enforcement,” Tasjar added. “There’s an incredible amount of information that can be gleaned without people’s consent.”

A National Institute of Standards and Technology study published in December 2019 found higher rates of false matches for Asian and African Americans, among other problems.

Tasjar noted that people often won’t be consenting to their faces being run through the facial recognition software, and therein lies another problem.

“I’d be shocked if this program were approved and if in two years it’s only limited to booking photographs,” he said. “What’s more likely to happen is the system will eventually query a much larger dataset of individuals.”

5. What’s the timetable for the Chino Police Department fully implementing this software?

Chino police said the department hasn’t settled on what company to purchase the facial recognition software from, but it could cost anywhere from $5,000 to $50,000 annually.

It also will require approval from the Chino City Council.  A date for the request to the council has not been set.


Source link