Fusion of Face Biometrics and Graph Neural Networks with Reinforcement `Learning for Secure Financial Fraud Detection.

– This research proposes a fraud detection framework that integrates reinforcement learning with graph neural networks to identify fraudulent financial transactions in real time. The system models relationships among cardholders, merchants, and devices while analyzing transaction patterns over time. Through a reward-based learning process, it improves detection accuracy, reduces false alarms, and ensures fast processing suitable for large-scale payment networks. Unlike traditional models that struggle with imbalanced datasets and static community detection, the framework adapts dynamically to evolving fraud patterns by reorganizing suspicious communities as they change. It combines structural connections, temporal sequences, and contextual details such as product type or location to strengthen robustness against adversarial behavior. Fraudsters cannot easily mimic legitimate activity across all dimensions simultaneously. To further enhance security, the system incorporates face biometrics during identity verification. Overall, the proposed approach delivers a scalable, adaptive, and secure solution for modern financial institutions.