Aware Ranks Top Performer in NIST Face Analysis Technology Evaluation (FATE) Benchmarking Test While Optimizing Demographic Parity

Aware’s facial presentation attack detection (PAD) algorithms achieved top ranking among 82 tested systems in the newly introduced NIST Face Analysis Technology Evaluation (FATE) test - NIST IR 8491.

BURLINGTON, Mass., Oct. 11, 2023 (GLOBE NEWSWIRE) -- Aware, Inc. (NASDAQ: AWRE) continues to build on its 30-year history of innovation, broad IP and biometric leadership and its commitment to deliver customers the highest level of technology with accurate, non-biased, fast algorithms, as evidenced by its performance in the National Institute for Standards and Technology’s (NIST) independent Face Analysis Technology Evaluation (FATE). The independent NIST results show Aware as a leader in presentation attack detection liveness technology while affirming Aware’s leadership position in optimizing demographic parity.

According to NIST, the FATE “evaluated the ability of software-based algorithms (software only, no hardware dependance) to detect the presence of presentation attacks in conventional 2D still and video imagery of human faces. The report surveyed the trade-off between false alarm rates on legitimate non-attack images and detection rates on images of various presentation attack instruments (PAIs).” Consistent with NIST’s Face Recognition Vendor Test (FRVT), the test results of NIST FATE are likely to become the trusted standard globally. This will bring a much-needed standard in benchmarking based on ISO 30107 (which is also called “anti-spoofing” or “liveness detection” for the solutions aimed at solving the presentation attack detection (PAD) problem).

As stated in the report: “The test results quantified the accuracy of passive, software-based face PAD algorithms applied to imagery of different PAIs… FATE PAD had two separate tasks that evaluated algorithms with an ability to detect 1) impersonation attacks and 2) evasion attacks... All algorithms [82] implemented the impersonation detection task, and 27 of those algorithms also implemented the evasion detection task... [The] evaluation dataset contained imagery (primarily stills) of approximately 20,000 presentation attack attempts and 21,000 bona fide presentations, which included 9 categories of presentation attacks.”