Spoofed Fingerprint Image Detection Using Local Phase Patch Segment Extraction and a Lightweight Network

• Published in: Advances in Digital Forensics XVIII pp. 85 - 105
• Main Authors: Abdullahi, Sani Mohammed, Sun, Shuifa, Malik, Asad, Khudeyberdiev, Otabek, Basheer, Riskhan
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing
• Series: IFIP Advances in Information and Communication Technology
• Subjects: Fingerprint spoofing detection; local phase quantization; neural network
• ISBN: 9783031100772; 3031100778
• ISSN: 1868-4238; 1868-422X
• DOI: 10.1007/978-3-031-10078-9_5

Fingerprint spoofing is one of the most successful attacks on fingerprint biometric systems. It involves the presentation of a fake fingerprint to a biometric sensor, which recognizes it as the original template and consistently uses it to authenticate an impostor as the genuine owner of the template. This chapter presents a methodology for combating fingerprint spoofing that employs local phase patch segment extraction and a lightweight triple-dense network. The methodology segments an input fingerprint image using local phase patch segment extraction, which also assists in extracting texture information so that each segment contains a consistent number of patches and each patch contains adequate minutiae information. The segmented image is fed to the lightweight triple-dense network, which is designed to generate discriminative information that helps distinguish between live and spoofed fingerprint images. This ensures optimum recognition accuracy and fast processing time while eliminating overfitting. Experimental evaluations using the LivDet 2013 and LivDet 2015 fingerprint datasets reveal that the methodology accurately classifies live and spoofed fingerprint images with an overall accuracy of 95.5%. Intra-class variation and inter-class similarity are eliminated by generalization without any accuracy degradation.