Biometric presentation attack detection shows maturity in NIST evaluation

The effectiveness of the best presentation attack detection (PAD) software for face biometrics is sufficient to give businesses that use it confidence they are catching most spoof attempts, at least based on the latest laboratory testing.

The inaugural evaluation of PAD technologies by the U.S. National Institute of Standards and Technology (NIST) for the Face Analysis Technology Evaluation (FATE) has results several of the leading developers can claim as wins. Perhaps most encouraging for businesses is the improved performance that can be achieved by combining multiple PAD algorithms.

The report shares evaluations of 82 passive, software-based face PAD algorithms operating on conventional 2D images. Each algorithm is submitted for evaluation in preventing impersonation, evasion or both, though in practice, all were implemented for impersonation, and 27 for evasion. The results are broken down into 9 attack types, represented by 16 tables for attacks with still media, and another for stills versus videos. Each is divided into convenience, measured as attack presentation classification error rate (APCER) at bona fide presentation classification error rate (BPCER) 0.01, and security, which is BPCER at APCER 0.01.

ID R&D, Alice Biometrics, Rank One Computing, South Korea-based Kakao Brain, iProov, CyberLink, Onfido, Aware, Saudi Arabia’s STCon and Thai developer Kasikorn Labs (part of Kasikorn Bank Group) each took the top spot in at least one category among the results tables. STCon, CyberLink, Alice, ID R&D, Kakao and ROC.ai each topped both the security and convenience sides of at least one results table.

ROC.ai did not register any false positives or negatives in the evasion test for presentation attack type 6, in one of several encouraging scores.

“Aware applauds NIST for both the rigor and the thought leadership demonstrated in designing this evaluation,” says Dr. Mohamed Lazzouni, CTO of Aware. “Aware strongly believes in building systems in tune with the operational conditions we encounter in our daily lives. As such, providing an optimal experience to balance security and convenience is critical. This is why Aware participated in both impersonation and evasion tasks and performed exceptionally well in both.”

The report also breaks down the datasets used, how long the algorithms take, and the results for different demographic groups.

Article Topics

biometric testing  |  biometrics  |  Face Analysis Technology Evaluation (FATE)  |  face biometrics  |  NIST  |  presentation attack detection  |  spoof detection