Stand-Alone Multi-Attention Fusion Network for Double-Identity Fingerprint Detection

In this work, we propose SAMAF, a novel stand-alone multi-attention fusion network to detect double-identity fingerprint images. The double-identity fingerprint is a newly developed potential threat where a fake, synthetic fingerprint is created by carefully aligning ridge-lines of two genuine fingerprints belonging to the criminal and innocuous accomplice. These fake fingerprints are visually indistinguishable due to smooth orientation fields at intersection regions and thus simultaneously exploited by both entities to evade the security system. The proposed SAMAF network stems from the novel multi-instance attention module to aggregate local and global discriminative contexts. On other hand, the multi-attention fusion module is designed to consolidate attentive features across multiple scales and focus on important regions. Experimental results and ablation studies on multiple fingerprint-based presentation attack datasets show the superiority of our proposed network as compared to state-of-the-art methodologies.