Iris liveness detection using texture segmentation

In order to chisel an iris recognition system and pass in the iris liveness test, an attacker can create semitransparent contact lens spoofing. Such a contact lens is transparent around the iris center and has fake iris texture of another person printed around the outer region. In this paper, such fake iris images are synthetically created which can obviate sleuthing even after using pupillary light reflex technique. Thereafter, new liveness detection is proposed to determine the perceptually invisible boundary between the fake and original iris textures. Response of Gaussian derivative filters with multiple scales and orientations at each pixel location is clustered using K-means to ascertain regions with different textures. To give robustness to the algorithm, it is iterated a certain number of times and a threshold mechanism is imposed to find the correct boundary. The proposed method is shown to achieve high liveness performance by generating a set of 600 fake iris images from the UPOL iris database.