Anti-spoofing detection based on eyeblink liveness testing for iris recognition

Abstract

Iris pattern is the most stable biometric trait for personal identification. It is the only traits that can’t be used after a person death. Despite its stability and difficulty to spoof, it has found that presenting a high quality image of an iris one can spoof and gain an access. Furthermore, the use of video frames of an authorized personnel and the use of 3D models can cheat the system. This study aimed at presenting a solution to this problem by testing the liveness of an eye being scanned by an access control device. The algorithm works by additional process of detecting an eyeblink and background subtraction and correlation to assume liveness. For one to gain access, first an iris is scanned and identified, secondly if this iris is in the database before providing an access, an eyeblink is also sensed. If eyeblink is sensed an access is granted otherwise access is denied. An algorithm has been developed in MATLAB adopting an adaptive Canny method for edge detection. The proposed algorithm validates the user being scanned by two stages which are; Eyeblink detection, background subtraction and correlation. Testing on standard datasets of ZJU Eyeblink, ACASIA v3 and the TalkingFace databases showed show 96.47% accuracy.