Fingerprint Spoof Detection: Temporal Analysis of Image Sequence

TL;DR

This paper presents a deep learning approach using CNN-LSTM to analyze fingerprint image sequences, leveraging temporal dynamics to effectively distinguish real fingerprints from spoofs, improving cross-material detection performance.

Contribution

The authors introduce a novel end-to-end deep learning method combining CNN and LSTM for temporal analysis of fingerprint sequences, enhancing spoof detection accuracy and generalization.

Findings

Improved cross-material TDR from 81.65% to 86.20% at FDR=0.2%.

Effective use of temporal features enhances spoof detection.

Large database with diverse spoof materials supports robust evaluation.

Abstract

We utilize the dynamics involved in the imaging of a fingerprint on a touch-based fingerprint reader, such as perspiration, changes in skin color (blanching), and skin distortion, to differentiate real fingers from spoof (fake) fingers. Specifically, we utilize a deep learning-based architecture (CNN-LSTM) trained end-to-end using sequences of minutiae-centered local patches extracted from ten color frames captured on a COTS fingerprint reader. A time-distributed CNN (MobileNet-v1) extracts spatial features from each local patch, while a bi-directional LSTM layer learns the temporal relationship between the patches in the sequence. Experimental results on a database of 26,650 live frames from 685 subjects (1,333 unique fingers), and 32,910 spoof frames of 7 spoof materials (with 14 variants) shows the superiority of the proposed approach in both known-material and cross-material (generalization) scenarios. For instance, the proposed approach improves the state-of-the-art cross-material performance from TDR of 81.65% to 86.20% @ FDR = 0.2%.