In 2014, the National Institute of Standards and Technology (NIST), in collaboration with the FBI, started studying how effective tattoo-recognition systems are, using tattoo images collected from inmates and arrestees to create a database that acts as a standardized metric. With the research set to expand dramatically, privacy advocates have raised serious concerns over not just how the research was conducted, but how law enforcement might use it in the future.

In a lengthy report, the Electronic Frontier Foundation lays out several apparent issues with NIST’s Tattoo Recognition Technology study. It used 15,000 FBI-provided images of tattoos from prisoners to create a data set, which the FBI then distributed to 19 outside parties. Those organizations, which ranged from universities to research institutions to private companies (including, EFF notes, MorphoTrak, which provides biometric tech to law enforcement agencies), applied their algorithms to data set, and reported back their effectiveness at identifying tattoos, identifying the same tattoo over time, and “matching common visual elements between tattoos” to help determine connections among individuals.

For its part, NIST maintains that because the study uses only isolated images, without accompanying information about the individuals from whom they came, it was not subject to the stringent regulation that would accompany human research studies.

“The project has been reviewed and determined to not meet the criteria for human subjects research as defined by federal regulations,” the organization said in a statement.

Tattoos, though, often don’t need accompanying information to help identify individuals. They can contain names, dates of birth, and other unique signifiers. More importantly, they can be used to identify someone’s religious, political, social, or other affiliation information that law enforcement can use to profile people, regardless of their actions.

NIST spokeswoman Gail Porter acknowledged the privacy concerns. “Once we were notified that a limited number of images could potentially contain identifiable information we notified the FBI,” she tells WIRED.

Like a Tattoo

Earlier this week, at Cannes, a photograph of Iranian actress Taraneh Alidoosti revealed an arm tattoo that appeared to be a feminist symbol. It caused a minor furor in her home country; Iran is not exactly a bastion of women’s rights. An image of her tattoo branded her, bringing down criticism from the traditionalist establishment.

Alidoosti’s tattoo was discovered because she was a public figure at a public event, and being a feminist is no crime in most parts of the world. Still, the incident’s a helpful reminder that tattoos have meanings beyond themselves. Replace the French paparazzi with law enforcement surveillance cameras, and a famous actress with a South Side teen, and it’s not hard to imagine the various ways in which tattoo recognition could be misused by law enforcement.

Police have long collected tattoo images from arrestees, just as they do fingerprints. There’s a difference, though, between maintaining a text-based database or physical photo albums, and a digital tattoo database that can be used to establish connections between individuals.

“What extent do they use this technology to put people in law enforcement databases and watch lists?” asks EFF investigative researcher Dave Maass. “If a tattoo recognition algorithm says a tattoo might be affiliated with gang activity, that person might be affiliated.”

Mass notes also that it’s not simply that tattoos could be used to identify such affiliations; they could also be used to misidentify. In one extreme example, Maass points out that a Chicago gang shares a six-pointed star symbol with the Jewish faith. Overlaps like that are not infrequent. “[Law enforcement] may end up roping in a lot of people who aren’t affiliated.”

Another potential issue? Tattoos can last longer than the associations they represent. If an individual doesn’t have the time, money, or inclination to remove a tattoo from a part of their life from which they’ve moved on, they could still be considered a party to it if a camera catches their ink.

Ultimately, Maass says, any concerns you may have about facial recognition can be applied to tattoo recognition as well.

Undersight

The EFF’s other major concern is the manner in which the research was conducted.

A major part of the Tattoo Recognition Technology initiative has been the Tattoo Recognition Technology Challenge (Tatt-C), which compiled those 15,000 images and distributed them to 19 third-party organizations.

“The Tatt-C competition required participants to perform a series of tests and report their results to NIST,” the EFF writes. “These experiments included identifying whether an image contained a tattoo and whether algorithms could match different images of the same tattoo taken over time.”

The research was also conducted outside of the Common Rule, a policy that mandates heightened oversight of experiments involving human subjects. That means there was no involvement from an Independent Review Board (IRB), which would have included either a prisoner or their representative to make sure ethical violations were not taking place.

“I don’t think researchers should be using prisoners as a free data resource,” says Maass. “In this country we have rules that are meant to protect prisoners, because of a long history of ethically challenged scientific research on them. If this study is going to use prisoner information, it needs to go through the highest possible scrutiny to make sure that people’s rights aren’t violated, and that prisoners know what has happened to them.”

The EFF’s position is that those 19 third parties should return the images they received, and destroy any back-ups. From there, it says, any further studies along this line should be overseen by an IRB.

For its part, NIST has already removed images of tattoos that contain information that could identify individuals from the Internet, and relayed its concerns to the FBI.

The next phase of the program, the Tattoo Recognition Technology Evaluation (Tatt-E), will use third-party algorithms to analyze a dataset of 100,000 tattoo images. They’ll be collected from law enforcement agencies in Florida, Michigan, and Tennessee. It had been scheduled for this summer, but is currently on hold.

“NIST will not run the Tatt-E evaluation until it has carefully reviewed the EFF’s report,” says Porter, “and performed a human subjects review of the project.”

Jump to original: 

The FBI’s Studying Prisoner Tattoos—And Pissing Off Privacy Hawks