A personalized benchmark for face anti-spoofing

Thanks to their ease-of-use and effectiveness, face authentication systems are nowadays ubiquitous in electronic devices to control access to protected data. However, the widespread adoption of such systems comes with security and reliability issues. This is because spoofs of face images can be easily fabricated to deceive the recognition systems. Hence, there is a need to integrate the user identification system with a robust face anti-spoofing element, which has the goal to detect whether a queried face image is a spoof or live. Most contemporary face anti-spoofing systems only rely on the query image to accept or reject tentative access. In real-world scenarios, however, face authentication systems often have an initial enrollment step where a few live images of the user are recorded and stored for identification purposes. In this paper, we present a complementary approach to augment existing face anti-spoofing benchmarks to account for enrollment images associated with each query image. We apply this strategy on two recently introduced datasets: CelebA-Spoof and SiW. We showcase how existing anti-spoofing models can be easily personalized using the subject's enrollment data, and we evaluate the effectiveness of the enhanced methods on the newly proposed datasets splits CelebA-Spoof-Enroll and SiW-Enroll.