Published in


California Police Are Sharing Facial Recognition Databases to ID Suspects

New emails reveal how a private technology company is working to change law enforcement as we know it

A composite image sourced from DataWorks Plus documents

MMany of California’s local law enforcement agencies have access to facial recognition software for identifying suspects who appear in crime scene footage, documents obtained through public records requests show. Three California counties also have the capability to run facial recognition searches on each others’ mug shot databases, and others could join if they choose to opt into a network maintained by a private law enforcement software company.

The network is called California Facial Recognition Interconnect, and it’s a service offered by DataWorks Plus, a Greenville, South Carolina–based company with law enforcement contracts in Los Angeles, San Bernardino, San Diego, San Francisco, Sacramento, and Santa Barbara.

Currently, the three adjacent counties of Los Angeles, Riverside, and San Bernardino are able to run facial recognition against mug shots in each other’s databases. That means these police departments have access to about 11.7 million mug shots of people who have previously been arrested, a majority of which come from the Los Angeles system.

An email from DataWorks Plus executive vice president Todd Pastorini offering details on California cities that use the company’s facial recognition services. Contact information has been redacted by OneZero.

Sacramento, Santa Barbara, and San Francisco also use the service, bringing the total number of mug shots in DataWorks Plus’ California system to 15 million, according to an email sent to the San Bernardino Sheriff’s Department from Todd Pastorini, executive vice president and general manager of DataWorks Plus. Though the public records request returned emails suggesting these three cities would be added to the sharing network, DataWorks Plus now tells OneZero that they’re not connected, meaning other cities cannot access their images, and vice versa.

Another email from Pastorini confirming that the Sacramento County Sheriff’s Department has joined the company’s network of facial recognition services. Contact information has been redacted by OneZero.

DataWorks Plus’ Interconnect network puts the company in a powerful position in the nation’s largest state. If police or sheriff’s departments invest in DataWorks Plus’ facial recognition system over a competitor’s, they could opt into having access to data from other cities around the state as well. Each contract is worth hundreds of thousands of dollars: When San Bernardino first bought DataWorks Plus’ technology in 2012, it financed the purchase through a $332,520 grant from the Department of Justice. The city renewed and upgraded its technology in July 2018 for $222,300.

But DataWorks Plus is also operating far beyond California. In 2017, it made a proposal to the Detroit police that listed 27 local, state, and federal agencies using the company’s facial recognition services, and company representatives call DataWorksPlus the number one provider of facial recognition on the West Coast. The Detroit proposal references Los Angeles’ adoption of the technology — by far the biggest in California — and boasts that the LAPD can search 7 million facial templates in less than 15 seconds.

An excerpt from DataWorks Plus’ Detroit proposal

The capabilities of the Interconnect system are detailed in that proposal, which outlines how a shared database could work with the Michigan State Police (MSP).

“If allowed by MSP, DataWorks Plus is can [sic] provide dual search capability — analysts will be able to select a probe and search Detroit and MSP databases from a single application,” the report read, emphasis included. That means in one piece of software, an analyst running facial recognition searches can search not only their own database, but the databases of other agencies as well.

By signing a contract with DataWorks Plus, Detroit police would be able to access records, including mug shots, from other locations.

DataWorks Plus also offered “an opportunity through mutual interest to access” databases in Chicago, Pennsylvania, New York, Northern New Jersey, Virginia, Maryland, and Columbus, Ohio. In other words, by signing a contract with DataWorks Plus, Detroit police would be able to access records, including mug shots, from those other locations. Pastorini says that this capability is not available to California law enforcement but did not elaborate on why.

Facial recognition is a contentious issue, especially in California. San Francisco and Oakland have both banned the technology’s use by public agencies. San Francisco banned the technology due to privacy concerns, while Oakland cited studies that show racial bias in facial recognition systems sold by large technology companies like Amazon.

Pastorini tells OneZero that the characterization of facial recognition in the media is often unfair because it pits forensic tools against consumer-grade technology. He says that the automatic biometric identification technology his company sells does not rely on machine learning or deep neural networks, the way tech companies like Amazon or Microsoft do. Instead, he insists that the facial recognition technology sold by DataWorks Plus was developed specifically for forensic use.

“I can put all these engines [algorithms] side by side, and the Amazon searches are not the best forensic searches,” he says. Pastorini argues that the recent trend of banning facial recognition is unfortunate, since there wasn’t evidence of the technology being misused in San Francisco.

But NEC, a Japanese tech vendor from which DataWorks Plus sources facial recognition algorithms, says its technology does partially rely on neural networks — the method Silicon Valley companies have used for everything from recognizing people on Facebook to generating fake human voices in Google Assistant. These algorithms also put images through a number of different analyses specific to facial matching. One NEC technique estimates a 3D model of a face based on a 2D image and then tries to match the face at different angles, which could be useful to match a face captured at different camera angles.

DataWorks Plus doesn’t make facial recognition software itself; rather, it builds custom tools for law enforcement and provides access to tools to be used on police data, like facial recognition and fingerprint matching. The company typically sells access to three different algorithms, or engines, from NEC, Rank One, and Cognitec. The engines aren’t perfect. In the process of reviewing a new engine for a potential upgrade, an analyst for the San Bernardino Sheriff’s Department wrote that a newer version “is much more forgiving with glasses and with the angle of the face (eyes don’t need to be level).”

Performance also varies greatly based on the engine being used, according to emails obtained through the public records request.

“Personally, I’ve never had a ‘hit’ with the Cognitec engine, so it would have to be a drastic improvement to be able to outperform either NEC or Rank One,” another analyst wrote during the same upgrade process.

The overall accuracy of the system is also unclear, even after reviewing the proposal submitted to Detroit police outlining the technology. A chart in the proposal ranks one of the algorithm’s performances on the Labelled Faces in the Wild dataset, a facial recognition dataset meant for research and specifically not built to be an accuracy test for facial recognition products, a co-creator of the dataset recently told OneZero.

The California Facial Recognition Interconnect isn’t the only system that helps state agencies share biometric data. In 1986, California established the Cal-ID system, which allows counties with more than 1.5 million residents to share and analyze fingerprints across the state. The state of California also runs Cal-Photo, a database of 32 million driver’s license photos that was created in 2002 to help law enforcement agencies share facial images. These photos can be downloaded and used to run facial recognition searches against, Pastorini says.

The Electronic Frontier Foundation (EFF), a nonprofit focused on digital rights, has objected to the use of facial recognition on the Cal-Photo database and expanding access to driver photos. EFF noted that the California DMV told the DOJ that a facial recognition and photo-sharing system wasn’t possible under current state laws.

Though his department could download Cal-Photo images and upload them into DataWorks Plus, San Bernardino Lieutenant Scott Landen, who leads the county’s Cal-ID program and oversees the operation of facial recognition and other biometric searches, tells OneZero it doesn’t.

“The software is never used as a guarantee.”

He also notes that the technology is used only as an investigative tool, rather than as evidence.

“It is very useful in narrowing down a possible suspect lead in an investigation,” Landen says. “The software just gives a possible individual. Our deputies still need to vet out this lead with more investigation, such as interviews and further follow-up. The software is never used as a guarantee that it is for sure the person suspected of committing the crime.”

The undercurrents of the future. A publication from Medium about technology and people.

Recommended from Medium

Brand Name: CHUWI

Working Virtually is a Dance Between Humans and Technology

Mass Navigation: How Ford Is Exploring the Quantum World with Microsoft to Help Reduce Congestion

Amateur radio vanity looping — the anarchist recipe

Are Airpod Pro’s still worth it in 2021?

Best GPU’s for Gaming

Smart Glasses Don’t Have to Look so Bad

Move fast and fix things: building the best hurricane response process in 12 hours.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Dave Gershgorn

Dave Gershgorn

Senior Writer at OneZero covering surveillance, facial recognition, DIY tech, and artificial intelligence. Previously: Qz, PopSci, and NYTimes.

More from Medium

Mudley Project#2 Timebase Rebalance AI — Payoff#18

The advantages of deploying an NLU bot to Verizon 5G Edge with SignalWire and AWS Wavelength

IoT Trends 2022 — LPWA IoT goes Hyperscale

Doing Something Wonderful in 2022