Designing Efficient Spoof Detection Scheme for Face Biometric

摘要

Spoofing attacks provided by fake individuals are considered as a major interest on biometric systems. To implement a robust face biometric system, deploying a reliable anti-spoofing scheme is needed. The concentration of this study to organize the anti-spoofing technique is on overlapped face textures together with image quality assessment. Our proposed fake detection scheme applies double anti-spoofing solution to distinguish live and fake identities. Firstly, image quality assessment method is used to differentiate fake and real samples by comparing their quality. The fake detection method using overlapped histograms of LBP texture descriptor is then applied for those samples recognized as real to increase the robustness of the biometric system against unreliable quality of images. Proposed spoof detection method presents an effective strategy for detecting fake face samples for video and print attacks. Demonstration of results on public spoof databases clarifies the robustness of the proposed solution for face fake detection.