Face Spoof Detection from Single Image Using Various Parameters

Abstract

To detect duplication of identity during authentication of online payment on mobile or personal computer, the automatic face recognition is widely used now days. The biometric presentation attacks can be performed to gain access to these systems. It is performed by presenting the authorized person’s photo or video. Hence it is important to study the various face spoof attacks. Currently proposed face spoof detection techniques have less generalization ability as these are not considering all factors and do not detect the spoofing medium.The four features such as specular reflection, blurriness, chromatic moment and color diversity are used to analyze the image distortion. The different classifiers are trained for printed photo attack and video replay attack to differentiate between genuine and spoof faces. We also propose an approach to detect the spoofing medium by checking the boundary of the captured image during the photo attack and video attack and an approach to detect the blinking of eye for detecting liveness. It gives us high efficiency rather than existing methods.