An exploration of pre-processing approaches for iris spoof detectors
Palak Verma, Arvind Selwal, Deepika Sharma · 2022 International Conference on Computational Intelligence and Sustainable Engineering Solutions (CISES) · 2022
Iris recognition has fascinated the attention of several real-life applications to offer secured and reliable human authentication. However, these systems are exposed to a diverse range of security breaches that include direct spoofing of the sensing device or using some indirect means to gain illegal access to the authentication system. To mitigate these spoof attacks on the sensor module, an anti-spoofing sub-module is integrated alongside the sensor that intelligently classifies a presented iris trait as real or fake. With the emergence of machine learning algorithms, the problem of spoof detection has become comparatively easier, where a model learns from training samples to detect the liveliness of a given image. Building a spoof detection model involves three important steps namely: pre-processing, feature extraction, and a classifier. In this study, we propound a comparative analysis of pre-processing techniques, which are widely deployed in data-driven iris spoof detectors (ISDs). The presented image pre-processing techniques are explored based on two categories of ISD mechanism namely; hand-crafted features or deep learning (DL)-based approaches. Our study reveals that pre-processing techniques significantly improves the quality of raw images that leads to better prediction accuracy.