Explainable Ai Integration Blockchain Fraud Detection System For Secure Financial Transaction

Abstract

Financial fraud represents one of the most critical threats to global economic stability, with estimated annual losses exceeding USD 5.1 trillion worldwide. Traditional rule-based detection systems have proven inadequate against increasingly sophisticated attack vectors, including synthetic identity fraud, account takeover (ATO) attacks, and coordinated transaction laundering. This paper presents AIBChain, a novel framework that synergistically combines deep learning-based anomaly detection with a permissioned blockchain ledger to deliver real-time fraud detection with sub-100ms latency. Our architecture integrates a Graph Neural Network (GNN) module for transaction relationship mapping, a Long Short-Term Memory (LSTM) autoencoder for temporal pattern recognition, and an Explainable AI (XAI) layer powered by SHAP values for regulatory compliance. The immutable blockchain audit trail ensures tamper-proof forensic evidence. Experimental evaluation on three real-world datasets PaySim, IEEE-CIS Fraud Detection, and a proprietary banking dataset comprising 47 million transactions demonstrates that AIBChain achieves an F1-score of 0.9741, AUC-ROC of 0.9923, and reduces false positive rates by 67.3% compared to state-of-the-art baselines. Smart contract-mediated consensus reduces inter-bank fraud dispute resolution time from 72 hours to 8.4 minutes. Our results conclusively establish AIBChain as a production-ready solution for next-generation financial security infrastructure