Hyp-OC: Hyperbolic One Class Classification for Face Anti-Spoofing

Face recognition technology has become an integral part of modern securitysystems and user authentication processes. However, these systems arevulnerable to spoofing attacks and can easily be circumvented. Most priorresearch in face anti-spoofing (FAS) approaches it as a two-classclassification task where models are trained on real samples and known spoofattacks and tested for detection performance on unknown spoof attacks. However,in practice, FAS should be treated as a one-class classification task where,while training, one cannot assume any knowledge regarding the spoof samples apriori. In this paper, we reformulate the face anti-spoofing task from aone-class perspective and propose a novel hyperbolic one-class classificationframework. To train our network, we use a pseudo-negative class sampled fromthe Gaussian distribution with a weighted running mean and propose two novelloss functions: (1) Hyp-PC: Hyperbolic Pairwise Confusion loss, and (2) Hyp-CE:Hyperbolic Cross Entropy loss, which operate in the hyperbolic space.Additionally, we employ Euclidean feature clipping and gradient clipping tostabilize the training in the hyperbolic space. To the best of our knowledge,this is the first work extending hyperbolic embeddings for face anti-spoofingin a one-class manner. With extensive experiments on five benchmark datasets:Rose-Youtu, MSU-MFSD, CASIA-MFSD, Idiap Replay-Attack, and OULU-NPU, wedemonstrate that our method significantly outperforms the state-of-the-art,achieving better spoof detection performance.