US Government Used Prisoners to Train Tattoo Recognition Algorithms

Government scientists are working with the FBI to create an automated tattoo recognition system that uses federal and state prisoners as a "bottomless pool of free data," according to a review of documents uncovered by the Electronic Frontier Foundation.

Tattoos are expressions of individuality, and getting inked typically communicates something unique about a person's experiences, personality, or beliefs.

It's no secret that law enforcement has long used tattoos to profile and identify criminal suspects. But more recently, the FBI teamed up with scientists at the National Institute for Standards and Technology (NIST) to create a system capable of identifying, grouping and making inferences about people based on their tattoos—something the EFF says raises serious First Amendment concerns.

One slide from an NIST presentation titled "Why Tattoos?" makes clear how law enforcement would use the system, which was developed alongside a group of third-party companies, universities, and private research institutions, to profile people en-masse. The slide explicitly states that tattoos "suggest affiliation to gangs, subcultures, religious or ritualistic beliefs, or political ideology" and "contain intelligence; messages, meaning and motivation."

What's more, the government researchers failed to disclose that 15,000 tattoo images used in the initial experiment, called "Tatt-C," came from prisoners who had not consented to the experiment—apparently in violation of federal rules for ethical research.

The use of prisoners in scientific research comes with a number of caveats to protect inmates from exploitation. Specifically, the rules limit the kinds of experiments that can be performed and require oversight by an independent review board that must include representation from at least one inmate.

But according to the documents obtained by the EFF, the researchers failed to disclose the experiment's use of prisoners, initially saying only that the data was "operationally collected." NIST higher-ups only approved the experiment retroactively, under the impression that it didn't involve data from human subjects at all.

The data was also distributed to 19 different third party companies and organizations, including MorphoTrak, a massive biometric technology firm that has done business with various US agencies.

"NIST researchers are using tattoo images obtained from prisons and jails without questioning whether the experiments require enhanced oversight to comply with federal ethical rules regarding research on prisoners. Rather, they are treating inmates as a bottomless pool of free data," EFF researchers Dave Maas and Aaron Mackey wrote.

"Now, with NIST and the FBI on the precipice of a new, larger experiment that will use upwards of 100,000 tattoo images, officials must suspend any further research into tattoo recognition technology until they address the First Amendment, ethical, and privacy concerns EFF has identified," Maas and Mackey wrote.

The tattoo recognition system would add to a growing number of biometric databases held by the FBI, many of which contain people who have never been suspected of a crime, let alone charged or convicted.

In a recent public filing, the FBI revealed its desire to make its enormous facial recognition repository completely exempt from the Privacy Act, arguing that it should keep data indefinitely so it can identify and predict crimes in the future. That database was projected to contain more than 52 million face images by the end of last year, including millions of innocent Americans who have applied for visas or received federal background checks as a condition of their employment.

Update: This article has been updated to note that the Tatt-C experiment also distributed tattoo data to third party companies.