An Enhanced Generative Adversarial Network Model for Fingerprint Presentation Attack Detection

Fingerprint recognition systems have played a significant role in the field of biometric security in recent years. However, it is vulnerable to several threats which can put the biometric security system at a significant risk. Presentation attack or spoofing is one of these attacks which utilizes a fake fingerprint created with a fabrication material by an intruder to fool the authentication system. Development of new fabrication materials makes this spoof detection more challenging for cross materials. In this work, we have proposed a novel approach for detecting these presentation attacks using Auxiliary Classifier-Generative Adversarial Networks (AC-GAN). The performance of the proposed method is assessed in an open set paradigm on publicly available LivDet Competition 2013 and 2015 datasets. Proposed methodology achieves an average accuracy of 98.52% and 92.02% on the LivDet 2013 and LivDet 2015 datasets, respectively which outperforms the state-of-the-art methods.