DoD works to break biometrics paradigm

It’s the stuff of movies: a high-ranking military officer scans his iris, thumbprint or voice pattern to access the top secret briefing room. An intelligence analyst zooms in on drone footage taken from hundreds of meters away to identify a single face with 98 percent certainty. An investigator drops a piece of DNA evidence into a briefcase-sized machine and receives a matching suspect almost immediately.

The Defense Department wants to make these scenarios a reality.

“It’s a very interesting time, being in biometrics. In 2018, we’ll be testing five different biometrics,” Will Graves, deputy project manager and Chief engineer for biometrics at DoD said on Identity Governance month.

Facial Recognition

While facial recognition software is already becoming fairly commonplace, it’s primarily used at short ranges, like within airports and other buildings.

Advertisement

“When I worked at [the Homeland Security Department], at-a-distance was out to 30 feet. At DoD, it’s out to almost a kilometer now,” Graves said. “When we say ‘face-at-a-distance’ we’re using a telephoto lens, but it’s not that easy to throw a telephoto lens on a video capture.”

At that kind of distance, you have to compensate for any number of potential distorting factors. Atmospherics can distort an image, so the camera has to have a high framerate. And algorithms have to account for heat waves.

Graves said in a recent test at 500 meters, the heat from a helicopter takeoff was enough to prevent the system from recognizing the presence of a human face, much less identifying it. The human eye, however, was still able to pick out the face.

Contactless

Contactless biometrics can run the gamut from on-the-move facial and iris recognition to no-touch fingerprint identification. DoD wants to use them to control access to forward operating bases.

Contactless fingerprint identifications requires using high-resolution cameras — about 500 pixels-per-inch — to get an image of a person’s fingerprints. One of the reasons for this is that people often balk at touching fingerprint scanners in crowded public settings, like airports. Graves said DoD has already had some success in matching contactless fingerprints to on-the-move facial captures.

But on-the-move facial recognition presents its own challenges, because cameras rarely catch a straight-on image of a person’s face in natural settings.

“Historically, facial recognition started out from taking mugshots,” Graves said. “So most facial recognition algorithms are very good with a full-frontal face. With the movement to video and some of the social media things, we have to start looking at algorithms that can take off-angle faces and do recognition that way.”

Video Analytics

Graves said DoD wants to tap into the vast amounts of information available on the dark web and via social media, and video is the future of social media. So it wants to be able to scrape identifying information from ISIS propaganda videos or bomb-making tutorials.

Grves said this has raised some privacy concerns among prominent social media watchdogs, but that they shouldn’t be concerned.

“Because [members of ISIS are] not U.S. citizens, the DoD doesn’t extend the Privacy Act of 1974 like DHS does. DHS provides privacy protections to everybody. The DoD has a little different interpretation of that. So, if they’re not U.S. citizens, they don’t really have that privacy protection,” Graves said.

But U.S. citizens would be protected from DoD prying into their data.

Rapid DNA

Graves said the DoD is working with the University of Virginia to develop smaller, more efficient equipment for DNA analysis, so it can be processed in the field rather than being sent to a lab. Right now, he said they’ve gotten it down to about 10 pounds, although it’s not quite ready to fit into a backpack.

The new system also uses CD technology, taking advantage of centripetal force to reduce the amount of chemicals required for testing, and thus the amount of chemicals needing to be transported.

Voice Recognition

Graves said DoD is trying to improve its algorithms that match voice samples against a watchlist to improve its ability to pick single voices out of the crowd. And DoD wants the capability to do so in handheld capture devices.

Graves said this fits into a larger need to adapt to a more mobile environment.

Common Access Cards

Graves said DoD is moving away from just using biometrics for law enforcement purposes, and moving toward managing who can access information, forward operating bases, and other locations and material. One possible way to do this is replacing Common Access Cards with biometrics.

“They’ve tied everything to the CAC card, from our travel system to our timekeeping system,” Graves said. “It’s going to take a lot of effort to unwind the ball of twine. The CAC card is into everything. We even have a daily system that tracks your attendance, and you have to use your CAC card.”

He said there are already locations, like the Mark Center in Alexandria, Virginia, that use iris and fingerprint scans to control access. It wouldn’t be that difficult to implement in other locations.

“When we get our CAC card, they already take our iris and our fingerprints, so they already have the biometrics for that. So we can use that database,” Graves said.

The 30-Year Roadmap

Graves said DoD has a roadmap to look 30 years into the future with biometrics and how it wants to explore potential uses. And sometimes, that roadmap changes. Graves said that often happens when commercial technologies accomplish a goal faster than military research. One example is when ATM machines in Japan began using vein pattern analysis, which traces unique patterns in the veins of a person’s hand, as a security feature.

But they’re also aware that servicemembers who would be using these biometrics in the field already have large amounts of equipment they’re responsible for carrying. Graves doesn’t want to add to that growing burden, but take advantage of capabilities that already exist by acknowledging that there are likely already existing sensors in the field.

“In the past, the paradigm was you had a sensor, and you had an algorithm. We’re trying to break that,” Graves said.