Biometrics innovation focus moving beyond matching, Identity Week panels predict

The advances on the horizon in biometrics are mostly found in areas like process improvement and supporting technologies, judging by presentations during the first day of Identity Week in Washington, D.C.

In a panel discussion moderated by Paul Hunter of Craefte, Ben Smith of the FBI’s CJIS division credited researchers like those at CITER, represented on the panel by Stephanie Schuckers of Clarkson University, with advancing the field. CJIS shares what data it can, as a partner, with CITER, which generally means biometrics, but not any corresponding identity information.

Matthew Lightner of the U.S. Border Patrol noted the importance of iris biometrics to the safety of Border Patrol officers.

The U.S. Border Patrol covers areas between official border crossings, so its officers see people who are trying to hide their identity on a regular basis, and in some cases have put effort and forethought into it. Because of this, the Border Patrol gets thousands of “manual overrides” based on, for example, a match of iris biometrics and no match of fingerprints.

In some cases, the individual is simply a bricklayer, or has otherwise innocently degraded fingerprints. In some cases, they have been burned off to avoid identification.

“Iris is a lot more reliable,” Lightner says.

Tim Meyerhoff of Iris ID noted that the nature of latent fingerprints means that modality will always be valuable, but also that iris biometrics have come a long way since his company began supplying scanners to Clear.

Smith notes that talks are ongoing about what to do with records that have bad fingerprint data, or none at all. CJIS criminal records are required to be linked to fingerprints, so new legislation may be necessary to make any change.

Explainability is on the research agenda at CITER, according to Schuckers. Improving the explainability of biometric matches is important to making them useful in court. Another area where more research is needed is in demographic differentials for liveness detection, which has received far less focus than it has gotten for face biometrics matching. CITER is also working on template protection through hashing.

Another development on the horizon is the harmonization of the Department of Homeland Security’s biometric scanners, for which an industry day was held in July. Lightner says as part of the agency, Border Patrol submitted its needs, which start with multiple modalities. When the process will be complete is uncertain, but Meyerhoff points out that NGI was introduced in 2008, but took until 2020 to become official, “so I can wait a little longer.”

For contactless fingerprinting, Border Patrol has already carried out tests, and found the technology promising. “As far as enrollment,” Lightner says, “we’re not there yet.”

Face biometrics innovation sweats the details

Facial recognition error rates have decreased tenfold in the past decade, Arun Vemury of DHS noted in a later panel. Yevgeniy Sirotin of SAIC and the Maryland Test Facility noted that it has also found its way into most people’s pocket during the same stretch of time.

“As a tester, I really marvel at how far this technology has come,” he said.

Sirotin points to demographic differentials, human-algorithm teaming, and face-adjacent technologies like presentation attack detection and image quality assessment as the areas of focus in the immediate future.

Joel Brogan of Oak Ridge National Lab points out that the massive improvements are due to a series of incremental steps by researchers, recounting how pose normalization was considered in the 2018 paper “To Frontalize or Not to Frontalize.”

Chris Centamore of ROC notes that algorithm size has also decreased dramatically, enabling facial recognition at the edge. This means that now a security guard, for example, can have a matching device with a watchlist of a million people on it in his hand.

He also predicts that the technology will soon more from acceptance to being expected, pointing out that weapons detection is already uncontroversial.

John Splain of Biometrics Guru suggested that Digital Travel Credentials, and equivalents that eliminate documents for other use cases, are part of the next step in face biometrics.

Face detection can still be improved, Vemury says, and Sirotin notes the need to work on processing groups of faces together for practical purposes in airport scenarios.

Perhaps less obvious, but about to be influential, according to Brogan, are “Foundation Models,” another form of generative AI.

During the following presentation, Will Graves of OBIM noted that DHS, the FBI and Department of Defense all use different image quality algorithms, pending the implementing ISO/IEC 29794-5. The roadmap for that process includes reference implementation development and a joint project between DoD, DHS and CITER in the second and third quarter of 2025, with a user acceptance test tentatively set for the first half of 2026.

The National Academies study on facial recognition will also inform U.S. government policy.

More innovation is coming to biometrics, just probably not in the dramatic improvements to match rates that recent developments in AI have delivered.

Article Topics

biometric matching  |  biometrics  |  border security  |  CITeR  |  digital identity  |  digital travel  |  explainability  |  face biometrics  |  Identity Week  |  Identity Week 2024  |  iris biometrics  |  U.S. Government