Person-specific Face Spoofing Detection for Replay Attack Based on Gaze Estimation

Based on gaze estimation, we propose an effective person-specific spoofing detection method to counter replay attack using a noninvasive challenge and response technique. The points on the computer screen create the challenge, and the gaze positions of the user as they look at the computer screen form the response. Firstly, face identification is conducted to recognize identity. Secondly, gaze estimation model is trained for each subject by adaptive linear regression with incremental learning and used to predict gaze positions when user is looking at the computer screen. Finally, difference between predicted gaze positions and system point locations is used as fake score to evaluate the liveness of user. Our basic assumption is that a genuine access can be attacked by salient objects and follow them. Therefore, the lower the fake score is, the more probable the user is genuine. Experimental results show that proposed method obtains competitive performance in distinguishing replay attacks from genuine accesses.