Shape Analysis Based Anti-spoofing 3D Face Recognition with Mask Attacks

With the growth of face recognition, the spoofing mask attacks attract more attention in biometrics research area. In recent years, the countermeasures based on the texture and depth image against spoofing mask attacks have been reported, but the research based on 3D meshed sample has not been studied yet. In this paper, we propose to apply 3D shape analysis based on principal curvature measures to describe the meshed facial surface. Meanwhile, a verification protocol based on this feature descriptor is designed to verify person identity and to evaluate the anti-spoofing performance on Morpho database. Furthermore, for simulating a real-life testing scenario, FRGCv2 database is enrolled as an extension of face scans to augment the ratio of genuine face samples to fraud mask samples. The experimental results show that our system can guarantee a high verification rate for genuine faces and the satisfactory anti-spoofing performance against spoofing mask attacks in parallel.