Deep Convolutional Neural Network for Double-Identity Fingerprint Detection

Abstract

Automatic human recognition using ubiquitous fingerprint sensors is the most widely used modality in modern biometric based security systems. The double-identity fingerprint is a fake fingerprint created by aligning two fingerprints for maximum ridge similarity and then joining them along an estimated cutline such that relevant features of both fingerprints are present on either sides of the cutline. The fake fingerprint containing the features of the criminal and his innocuous accomplice can be enrolled with an electronic machine readable travel document and later used to cross the automated border gates by claiming identity of the accomplice. In this letter, we have developed a deep convolutional neural network (CNN)-based patch-learning approach to estimate the cutline by training the network to identify and learn the pattern around the region of the joint fingerprint. This is a recent, new fingerprint alteration technique, and due to the unavailability of any such public database, we have generated a new database of 450 double-identity fingerprints. Experimental results show that the deep learning based approach is able to predict the cutline with an equal error rate, which is the best when compared with many other popular handcrafted features for double-identity fingerprint detection.