Bridget Chinalu Ujah-Ogbuagu
Volume: 12 Issue: 04, 2024
Abstract:
The increasing reliance on cloud-based AI solutions for financial fraud detection introduces significant challenges related to data privacy, security, and computational efficiency. To address these challenges, we propose a Cloud-Enabled Federated Graph Neural Network (CE-FGNN) framework, integrating privacypreserving Federated Learning (FL) with Graph Neural Networks (GNNs) for enhanced fraud detection. The framework incorporates differential privacy techniques to safeguard sensitive banking data while leveraging the Firefly Algorithm (FA) for optimizing model convergence across distributed nodes. Our approach effectively mitigates risks associated with centralized data storage by ensuring secure, decentralized learning among financial institutions. Extensive experiments conducted on the Bank Account Fraud Dataset Suite (NeurIPS 2022) demonstrate that CE-FGNN outperforms state-of-the-art models. The framework achieves an accuracy of 98.76%, precision of 97.89%, recall of 98.34%, and F1-score of 98.11%, surpassing traditional FL-GNN models by an average of 3.7% across all metrics. Additionally, CE-FGNN reduces computational overhead by 22.5% compared to conventional centralized approaches, enabling scalable real-time fraud detection. The proposed framework significantly enhances fraud detection accuracy while maintaining strict privacy standards, making it a viable solution for secure AI-driven financial analysis. Future work aims to extend this approach to multi-modal financial datasets for broader applicability in banking security.