Optical Coherence Tomography-Based Fingerprint Presentation Attack Detection via Multi-Tissue Features Contribution Analysis

Currently, unsupervised anti-spoofing methods suffer from low accuracy and susceptibility to irrelevant factors. In order to solve these problems, this paper analyzes the contribution of different fingertip tissue structures to anti-spoofing tasks and proposes an unsupervised anti-spoofing method based on the weighted contribution of tissue structures. Unsupervised OCT anti-spoofing methods suffer from weak robustness and lack comprehensive exploration of sub-dermal structure features. The proposed method introduces quantified weights of fingertip tissue contributions with self-attention through the Shapley value. The module can amplify crucial fingertip features and extracts more key fingertip information, thereby improving the anti-spoofing performance.