Fingerprint Presentation Attack Detection Using A Novel Multi-Spectral Capture Device and Patch-Based Convolutional Neural Networks

We introduce a novel capture device that provides two new powerful sensing modalities for fingerprint presentation attack detection (FPAD): multi-spectral short-wave infrared (SWIR) imaging and laser speckle contrast imaging (LSCI). A prototype touchless fingerprint capture device was developed employing a 64×64 indium gallium arsenide (InGaAs) sensor and a high definition visible light camera. The former provides presentation attack detection capabilities while the latter enables backward compatibility with legacy fingerprint matching systems. A patch-based convolutional neural network (CNN) classification algorithm was designed to classify patches from SWIR and LSCI data as skin or non-skin. The final decision is made for a given fingerprint presentation based on the average classification score of individual patches. The classifiers for the two sensing modalities demonstrated excellent performance on a diverse dataset captured with our device, proving the power of SWIR and LSCI imaging for FPAD.