A Novel Face Spoofing Detection Method Based On Gaze Estimation

Since gaze is a kind of behavioral biometrics which is difficult to be detected by the surveillance due to the ambiguity of visual attention process, it can be used as a clue for anti-spoofing. This work provides the first investigation in research literature on the use of gaze estimation for face spoofing detection. Firstly, a gaze estimation model mapping the gaze feature to gaze position is established for tracking user's gaze trajectory. Secondly, gaze histogram is obtained by quantifying and encoding the gaze trajectory. Finally, information entropy on gaze histogram suggests the uncertainty level of user's gaze movement and estimates the liveness of the user. Our basic assumption is that the gaze trajectory of genuine access has higher uncertainty level than that of attack. Therefore, the greater the entropy, the more probable the user is genuine. Experimental results show that the proposed method obtains competitive performance in distinguishing attacks from genuine access.