On the importance of using high resolution images, third level features and sequence of images for fingerprint spoof detection

Abstract

The successful and widespread deployment of biometric systems brings on a new challenge: the spoofing, which involves presenting an artificial or fake biometric trait to the biometric systems so that unauthorized users can gain access to places and/or information. We propose a fingerprint spoof detection method that uses a combination of information available from pores, statistical features and fingerprint image quality to classify the fingerprint images into live or fake. Our spoof detection algorithm combines these three types of features to obtain an average accuracy of 97.3% on a new database (UNESP-FSDB) that contains 4,800 images of live and fake fingerprints. An analysis is performed that considers some issues such as image resolution, pressure by the user, sequence of images and level of features.