Rumors of liveness detection’s defeat have been greatly exaggerated

Photo and video face filters are perhaps the most mainstream use case for augmented reality –  and an illustrative test case for biometric facial recognition. Results from a series of tests run by Corsight AI demonstrate the power of its Facial Intelligence platform when presented with heavily filtered images popular with users of TikTok, Instagram and other social media sites.

Face filters that added facial hair, doggie ears and radical distortions all failed to fool Corsight’s facial recognition technology. According to a press release, so-called “funny filters” were chosen to gauge whether the system could still match an altered face with the correct individual from a database of 100,000 faces. Facial Intelligence scored a cool one hundred percent, accurately recognizing the person and correctly identifying them as the most probable match in every instance tested.

For Corsight AI, the results are significant as a demonstration of its platform’s ability to make accurate matches under challenging conditions. Being able to discern true identities behind severe digital distortions speaks to Facial Intelligence’s potential for use in law enforcement and military scenarios, where low-quality video, restricted camera angles, and scenarios involving darkness, motion, or disguise are common obstacles to identifying criminals. Per the release, “it becomes invaluable for law enforcement in searching for a wanted or missing person, or for retail security in real-time identification of a known shoplifter, even when the reference image is outdated, of low quality, or taken from a challenging angle.”

FaceTec detects a falsehood in spoofing video

FaceTec is calling out a new tactic in the game of spoofs: a video that falsely portrays its liveness tool as being vulnerable to its attack. Posting on LinkedIn, FaceTec CEO Kevin Alan Tussy says the video in question, for the Volcam camera spoofing tool, shows his company’s software FaceTec to imply “that their injection software works against FaceTec, but it doesn’t. Of course, they end the video before you can see that our AI handily rejects the attack, but because FaceTec is the biggest, they want to try to leverage us. Scammers scamming Scammers.”

Call for papers on biometric authentication safeguards

Researchers from BioID, the University of Bologna and the University of Magdeburg are seeking articles for a collection to be published in the journal Frontiers in Imaging. A post on LinkedIn says that the curators “expect papers that explore the latest advances in biometric authentication and security.” Both review and research papers are welcome.

Specific topics of interest to the project include face, fingerprint and iris biometrics; voice morphing and injection attacks; bias in biometric matching systems; deepfake detection; model poisoning attacks on neural network-driven biometric systems; and vulnerabilities in KYC, among others.

The results, say the guidelines, “should be useful for developers of commercial biometric systems, highlighting the potential weaknesses in their systems and providing them with a path to more advanced and secure biometric user authentication solutions.”

Guidelines, deadlines and other information can be found here.

Article Topics

BioID  |  biometric liveness detection  |  biometrics  |  Corsight  |  face morphing  |  FaceTec  |  injection attacks  |  spoofing