Real-Time Fraud Detection Using Graph Neural Networks and Federated Learning
DOI:
https://doi.org/10.63282/3050-9246/ICRTCSIT-103Keywords:
Fraud Detection, Graph Neural Networks, Federated Learning, AI, Banking, Privacy, Cybersecurity, Real-Time Systems, GDPR, Anomaly DetectionAbstract
Fraud continues to be one of the most pressing threats to financial systems, costing banks and consumers billions annually. Traditional fraud detection methods rely heavily on rule-based engines and static anomaly detection, which struggle against adaptive adversaries and organized fraud rings. This paper introduces a hybrid framework that leverages Graph Neural Networks (GNNs) for relational fraud detection and Federated Learning (FL) for privacy-preserving model collaboration across financial institutions. GNNs enable the identification of hidden connections between accounts, devices, and transactions that traditional models often overlook, making them highly effective against collusive schemes. Federated learning allows institutions to train models jointly without exposing sensitive customer data, aligning with privacy and regulatory requirements such as GDPR and CCPA. We present a reference architecture for federated GNN-based fraud detection, with components for local graph construction, secure aggregation, and global model dissemination. Case studies demonstrate improvements in fraud detection accuracy, reductions in false negatives, and enhanced resilience to evolving attack patterns. Challenges such as communication overhead, interpretability of GNN decisions, and cross-border regulatory acceptance are discussed. We also highlight future research directions, including quantum-accelerated graph learning and generative adversarial fraud simulations. Our findings suggest that the combination of GNNs and FL is a transformative step in building real-time, scalable, and secure fraud detection systems for global banking
Downloads
References
[1] W. Hamilton, R. Ying, and J. Leskovec, 'Representation Learning on Graphs: Methods and Applications,' IEEE Data Eng. Bulletin, 2017.
[2] B. McMahan et al., 'Communication-Efficient Learning of Deep Networks from Decentralized Data,' AISTATS, 2017.
[3] A. Kumar and S. Ghosh, 'Graph Neural Networks for Fraud Detection in Financial Systems,' IEEE Access, 2023.
[4] Google AI, 'Federated Learning: Collaborative Machine Learning without Centralized Training Data,' 2017.
[5] S. Raza et al., 'Secure Aggregation for Federated Learning in Finance,' ACM CCS, 2021.
[6] J. Chen et al., 'FastGCN: Fast Learning with Graph Convolutional Networks via Importance Sampling,' ICLR, 2018.
[7] World Economic Forum, 'Future of Financial Infrastructure: Blockchain and AI,' WEF Report, 2020.
[8] S. Sun and Q. Li, 'Adversarial Attacks on Graph Neural Networks,' arXiv preprint, 2019.
[9] NIST, 'Privacy Framework for Data Protection,' NIST Publication, 2020.
[10] Accenture, 'AI in Fraud Detection: Global Banking Trends,' Accenture Research, 2022.
[11] KPMG, 'Federated Learning in Financial Services,' KPMG Report, 2021.
[12] IBM Research, 'Graph AI for Enterprise Fraud Detection,' IBM Whitepaper, 2022.
[13] Microsoft Azure, 'Scalable AI for Financial Fraud Prevention,' Azure Report, 2023.
[14] OECD, 'AI and Financial Crime Prevention,' OECD Report, 2021.
[15] McKinsey & Co., 'The Next Frontier in Fraud Detection,' McKinsey Insights, 2023.
[16] Thirunagalingam, A. (2024). Bias Detection and Mitigation in Data Pipelines: Ensuring Fairness and Accuracy in Machine Learning. Available at SSRN 5047605.
[17] Singhal, S., Kothuru, S. K., Sethibathini, V. S. K., & Bammidi, T. R. (2024). ERP excellence a data governance approach to safeguarding financial transactions. Int. J. Manag. Educ. Sustain. Dev, 7(7), 1-18.
[18] Praveen Kumar Maroju, "Assessing the Impact of AI and Virtual Reality on Strengthening Cybersecurity Resilience Through Data Techniques," Conference: 3rd International conference on Research in Multidisciplinary Studies Volume: 10, 2024.
[19] Sehrawat, S. K., Dutta, P. K., Bhatia, A. B., & Whig, P. (2024). Predicting Demand in Supply Chain Networks With Quantum Machine Learning Approach. In A. Hassan, P. Bhattacharya, P. Dutta, J. Verma, & N. Kundu (Eds.), Quantum Computing and Supply Chain Management: A New Era of Optimization (pp. 33-47). IGI Global Scientific Publishing. https://doi.org/10.4018/979-8-3693-4107-0.ch002
[20] L. N. R. Mudunuri, V. M. Aragani, and P. K. Maroju, "Enhancing Cybersecurity in Banking: Best Practices and Solutions for Securing the Digital Supply Chain," Journal of Computational Analysis and Applications, vol. 33, no. 8, pp. 929-936, Sep. 2024.
