Study: Pinellas sheriff's facial recognition system has danger of abuse and little oversight

The PInellas Sheriff's Office facial recognition database matches images similar to one from a suspect's fraudulent driver's license.
The PInellas Sheriff's Office facial recognition database matches images similar to one from a suspect's fraudulent driver's license.
Published Oct. 18, 2016

A study released today found that a facial recognition database managed by the Pinellas County Sheriff's Office and used by law enforcement agencies across Florida is subject to little oversight and lacks transparency.

According to the report, released by the Center on Privacy and Technology at Georgetown Law, the Sheriff's Office has broad guidelines outlining the use of the software that raise privacy concerns and could lead to misuse, the study's authors said.

"You are trackable by the police in a way that you really have never been before," said Alvaro Bedoya, the center's executive director.

But Sheriff Bob Gualtieri defended use of what he said is an efficient tool containing information that was either already public record, such as arrest photos, or that law enforcement already had access to, like driver's license photos. The agency hasn't experienced issues with misidentification or misuse, he said.

"This is probably one of the things where there's the least opportunity for abuse because all it is, is an analytical tool," he said.

The Sheriff's Office launched the database in 2001 with federal grant money, making it one of the longest-running programs of its kind, according to the study's authors. It has several applications when trying to identify someone, Gualtieri said. A deputy can, for example, run a search of someone who gives a false name, or use it to learn the identity of a dead person.

They can also use it if they have photographic evidence or video surveillance of a crime. The Sheriff's Office gave the example of a case out of Boynton Beach this year in which a man who used a fake driver's license to pick up two Apple MacBooks at a Best Buy was identified through his fraudulent ID photo.

The database is used by all of Pinellas County's roughly 800 deputies as well as officers from 243 agencies across the state, including the St. Petersburg and Tampa police departments and the Hillsborough and Pasco county sheriff's offices. Pinellas has also partnered with 40 of those agencies to expand the number of arrest photos in the database.

The Sheriff's Office has guidelines for its deputies that specify the software can be used only for official investigations and that deputies are encouraged to use the software "whenever practical." The guidelines state that deputies should get consent before taking a photo, unless a person is in a public space or the photo is taken off social media. Gualtieri emphasized that the software is never used for "random fishing expeditions" to acquire new photos.

The guidelines don't include a reasonable suspicion standard and don't require the Sheriff's Office to conduct audits to ensure deputies are using the software fairly, which "opens the floodgate for misuse and abuse," Bedoya said. The standards also apply only to Pinellas deputies. It is up to other agencies to come up with their own rules.

Keep up with Tampa Bay’s top headlines

Keep up with Tampa Bay’s top headlines

Subscribe to our free DayStarter newsletter

We’ll deliver the latest news and information you need to know every weekday morning.

You’re all signed up!

Want more of our free, weekly newsletters in your inbox? Let’s get started.

Explore all your options

Facial recognition systems have been found to have errors, Bedoya said. He pointed to a 2012 study from the Institute of Electrical and Electronics Engineers using Pinellas County's database which found that the algorithms had less accuracy identifying women, black people and those ages 18 to 30.

"The technology is powerful, but it's not neutral," Bedoya said.

The tool has been mainly used at the front end of investigations, not as evidence in the courtroom. Pinellas County Public Defender Bob Dillinger told the study's authors he wasn't aware of any disclosures in his office regarding use of the software after a crime, such as from surveillance video in a bank robbery. His lawyers had dealt only with disclosures involving deputies identifying people with no ID or who were giving fake names.

But he agrees with the study's authors that it leaves room for abuse, comparing it to cases in which officers have used the Driver and Vehicle Information Database on neighbors or ex-girlfriends.

"If it happened there, it can happen here in this facial recognition," he said.

The study includes suggested legislation that includes guidelines for officers to have reasonable suspicion before using facial recognition, require a warrant for a search of driver's license photos and to audit the systems to test them for accuracy — a suggestion Gualtieri called "totally unnecessary" and shows the study's authors have an agenda.

"This is doing no more than sorting things out in a more sophisticated way than a more manual way," he said. "It's not doing anything that affects or relates to anyone's privacy.

Contact Kathryn Varn at (727) 445-4157 or Follow @kathrynvarn.