Secure and Transparent Banking:explainable AI-Driven Federated Learning Model for Financial Fraud Detection

This paper presents a secure and transparent financial fraud detection framework that integrates Federated Learning (FL) with Explainable Artificial Intelligence (XAI). The proposed system enables multiple financial institutions to collaboratively train a global fraud detection model without sharing sensitive transaction data, thereby preserving privacy and supporting regulatory compliance. A deep neural network-based classifier is combined with the Federated Averaging (FedAvg) algorithm to identify complex fraud patterns in distributed environments. To improve transparency and trust, the framework incorporates LIME-based explanations and a hybrid rule-based decision mechanism that provides human-readable insights into suspicious transactions. The system also includes a real-time Streamlit-based monitoring interface for fraud prediction, risk visualization, and transaction analysis. Experimental results demonstrate that the proposed model achieves high fraud detection accuracy while maintaining privacy, interpretability, and scalability for practical banking applications.