A DCNN Based Fingerprint Liveness Detection Algorithm with Voting Strategy

The concern of the safety of fingerprint authentication system is rising with its widely using for it is easy to be attacked by spoof (fake) fingerprints. Fake fingerprints are usually made of Ploy-Doh, silicon or other artifacts. So most current approaches rely on fingerprint liveness detection as main anti-spoofing mechanisms. Recently, researchers propose to use local feature descriptor for fingerprint liveness detection, but the results are still not satisfying the real world application requirement. Inspired by the newly trend of application of Deep Convolution Neural Network (DCNN) in computer vision field and its outstanding performance in face detection and image classification, we propose a novel fingerprint liveness detection method based on DCNN and voting strategy, which performs better than handcraft feature and optimize the process of feature extraction and classifier training simultaneously. The experimental results on the datasets of LivDet2011 and LivDet2013 show that the proposed algorithm has great improvement compare to the former state-of-the-art algorithm, and keep highly real-time performance at the same time.