Tattoos aren't just for rebels: 1 in 5 American adults have some ink, according to recent polls. And now the government is trying to beef up technology that can automatically identify people by their tattoos.
The National Institute for Standards and Technology, a part of the Commerce Department that has taken the lead on evaluating biometrics, organized a "challenge" in which groups faced off to see who could deliver software and algorithms that identified tattoos most accurately. The event, sponsored by the FBI's Biometric Center of Excellence, brought together researchers from academia and the private sector to test image recognition technology against five different scenarios.
Groups of researchers, including some from Purdue University and the MITRE Corporation, worked with data sets ranging from hundreds to thousands of images. The results from six organizations were reviewed at a NIST workshop on Monday.
Some of the systems had "hit rates well above 90 percent" in some of the tests, like basic tattoo detection and identification over time as well as matching up a small piece of a tattoo to an image showing the the whole thing, said Mei Ngan, the NIST computer scientist who organized the event.
But other tests, like using sketches or computer graphics to match up with a real tattoo and trying to identify shared characteristics across different tattoo designs were more challenging.
Tattoos could help identify victims of disasters, like tsunamis or earthquakes. But tracking down criminals, who often sport tattoos at a higher rate than the general population, Ngan said, is the main reason the government is so interested in the technology.
"You can't use it as a primary biometric like a finger print or face because it's not necessarily uniquely identifying," Ngan said. "But it can really help in cold cases where you don't have those things."
Images of tattoos are already part of the FBI's massive Next Generation Identification database, which focuses on biometric identifiers, including fingerprint and facial recognition. But it currently relies on written descriptions of tattoos to help identify suspects, said Ngan.
But text descriptions are "inadequate," she said. Many people may choose similar motifs -- from crosses to brand logos -- for their tattoos. And even vastly different tattoos may be described with similar language.
"Think of a Hello Kitty tattoo versus a wildcat tattoo -- they both might be put in a text-based database as 'cat,'" Ngan said.
Tattoo identification may raise fewer privacy concerns than some other types of physical identifiers, according to Alvaro Bedoya, the executive director of Georgetown Law's Center on Privacy & Technology. "Tattoos are traditionally something you put on your body to distinguish yourself from other people -- it's a little different from someone recognizing your face because a digital description has been logged, or analyzing your gait," he said.
Still, Bedoya questions if the law has caught up with how image recognition technology will likely be used with tattoos: "People are being identified remotely without their knowledge -- and right now the Fourth Amendment doesn't really say anything about that."