Abstract

The increasing digitization of financial transactions has escalated the scale and sophistication of cyber-enabled financial fraud, rendering traditional rule-based defense mechanisms inadequate. This paper presents FraudSense AI, an autonomous defense model that leverages advanced artificial intelligence (AI), machine learning (ML), and graph-based network analysis to detect, predict, and mitigate financial cyber threats in real time. By modeling transaction flows and user interactions as a dynamic graph and applying a hybrid deep learning + anomaly detection approach, FraudSense AI adapts to evolving fraud patterns, identifies collusive fraud, synthetic identity fraud, insider threats, and triggers automated response workflows. Experiments on benchmark and simulated financial transaction datasets demonstrate the model’s high detection accuracy, low false-positive rate, and effective real-time responsiveness, outperforming traditional rule-based and classic ML-based fraud detection systems. FraudSense AI thus offers a scalable, adaptive, and resilient solution for modern financial institutions.