Liveness Detection Based on Human eye Blinking for Photo Attacks

Abstract

Auto face recognition mainly implemented to avoid the replication of identity to demonstrate through security check. This rage of face verification has brought intensive interest about facial biometric towards attacks of spoofing, in which a person’s mask or photo can be produced to be authorized. So, we propose a liveness detection based on eye blinking, where eyes are extracted from human face. The method of face recognition was applied by utilizing OpenCV classifier and dlib library, and a concept of edge detection and calculation of structure to extract the portion of the eye and to observe and make note of variation in the attributes of the eyes over a time period was employed. The landmarks are plotted accurately enough to derive the state of eye if it is closed or opened. A scalar quantity EAR (eye aspect ratio) is derived from landmark positions defined by the algorithm to identify a blink corresponding to every frame. The set of EAR values of successive frames are detected as a eye blink by a OpenCV classifier displayed on a small window when person is in front of camera. Finally, it gives the accuracy result whether it is human being or spoof attack.