Empirical Evaluation of Texture-Based Print and Contact Lens Iris Presentation Attack Detection Methods

Abstract

Iris-based identification methods have been popularly used in real-world applications due to the unique characteristics of iris when compared to other biometric characteristics like face and fingerprint. As technological advances and low-cost artefacts are becoming more available, vulnerabilities to iris biometrics due to presentation attacks (PAs) are becoming a challenging problem. Presentation attack detection (PAD) algorithms have been employed in biometric capture devices and it has been an active research topic in the past years. In this study, a detailed survey and evaluation of state-of-the-art texture-based iris PAD methods are performed. Five different PAD methods are tested on four different datasets consisting of print and contact lens presentation attacks. Extensive experiments are performed on four different scenarios of presentation attack and results are presented. The properties of PAD algorithms like the quality of the database, the generalization abilities are mainly discussed in this work. It has been observed that fusion-based PAD methods perform better than other methods.