Statistical texture analysis-based approach for fake iris detection using support vector machines

Abstract

This paper presents a novel statistical texture analysis based method for detecting fake iris. Four distinctive features based on gray level co-occurrence matrices (GLCM) and properties of statistical intensity values of image pixels are used. A support vector machine (SVM) is selected to characterize the distribution boundary, for it has good classification performance in high dimensional space. The proposed approach is privacy friendly and does not require additional hardware. The experimental results indicate the new approach to be a very promising technique for making iris recognition systems more robust against fake-iris-based spoofing attempts. © Springer-Verlag Berlin Heidelberg 2007.