Abnormal and Spoofing Iris Detection using Image Feature and Liveness

Abstract

Iris information is unique and measurable which is used to recognize the identity or verify the claimed identity of individual. Since this is sensitive information, identity verification using forgery iris data can cause serious problems in areas such as finance and terrorism. So, it must be done through a rigorous authentication system. In this paper, we deal with authenticating iris information and detecting various abnormal irides including as cosmetic contact lens, printed iris images, and counterfeit eyeballs. The proposed detecting abnormal iris algorithms consist of two major processes, Liveness detection and Abnormal iris detection. The liveness of iris image is evaluated by inspecting the variation of eyelid curvature during blink. Liveness test filters the most of counterfeit eyeballs and printed iris images. Then, the features of true iris images are examined to detect cosmetic contact lens and spoofing iris images which have passed the Liveness detection process. In Abnormal iris detection, we classify irides by creating the probability models using those features. According to the experiments, the proposed algorithms detect abnormal iris inputs with 92.22% accuracy. The proposed system prevents the various types of abnormal iris and can perform without the help of additional hardware devices. Thus, it makes iris authentication systems to be more reliable and useful.