Exploring Pre-Processing Approaches for Deep Learning-based Fingerprint Spoof Detection Mechanisms

Fingerprint-based human authentication being most widely deployed systems are also exposed to many security threats. Among all, spoofing attacks is widely attempted that involves circumventing the sensor module of the system by presenting a fake replica of the original trait of a genuine user. A counter measuring mechanism is deployed in these systems that intelligently measure the vitality characteristics of a presented fingerprint image. These sub-modules are integrated besides sensing device of the system and popularly known as anti-spoofing methods or fingerprint spoof detection (FSD) mechanisms. The evaluation of data-driven enabled computer vision paradigm has facilitated researchers to design intelligent FSD. Pre-processing data to train these FSD models is required to improve the overall quality of the input images. This paper presents a focused review of pre-processing approaches used in deep learning-based FSD models. The proposed research study has clearly observed that image pre-processing is a vital step to significantly improve the overall accuracy and efficiency of FSD approaches.