An AI Security Agent for Banking: Multi-Vector Fraud and AML Detection Across Retail and Corporate Accounts

The paper presents an AI security agent designed for banking that integrates a three-component fusion architecture to detect both signature-based fraud and behavioral financial crimes across retail and corporate accounts. It utilizes a dual-stream approach, combining LSTM sequence models, statistical monitors, and graph modules, achieving F1 scores of 0.787 for transaction streams and 0.867 for session streams, significantly outperforming traditional rule-based and LSTM-only baselines. This model is crucial for practitioners as it offers enhanced detection capabilities and real-time responses, addressing complex fraud schemes that evade conventional detection methods.