Hypergraph Neural Networks for Robust Fingerprint Matching in Forensic Applications

Abstract-Fingerprint matching is a crucial task in forensic science, where the accurate and reliable identification of individuals is essential for criminal investigations. Traditional fingerprint matching algorithms often struggle with challenges such as occlusion, distortion, and partial prints. In this study, we propose a novel approach that leverages Hypergraph Neural Networks (HGNNs) to enhance the robustness and accuracy of fingerprint matching in forensic applications. By modeling fingerprint features as hypergraphs, we capture higher-order relationships between minutiae points and their spatial configurations, enabling more effective matching despite partial or degraded fingerprints. The HGNN framework integrates both local and global feature information, improving the system’s ability to recognize subtle and complex patterns in fingerprint data. Extensive experiments on benchmark fingerprint datasets demonstrate that our approach outperforms conventional methods in terms of matching accuracy and robustness to noise. The proposed HGNN-based model provides a promising solution for advancing forensic fingerprint identification systems, offering improved performance under challenging real-world conditions.