Federated Learning Framework for Privacy-Preserving Cross-Bank Fraud Detection
DOI:
https://doi.org/10.63282/3050-9262.IJAIDSML-V6I1P126Keywords:
Federated Learning, Privacy-Preserving Machine Learning, Financial Fraud Detection, Cross-Bank Collaboration, Secure Aggregation, Differential Privacy, Byzantine-Robust Aggregation, Data Privacy Regulations, Distributed Machine Learning, Banking Transaction Security, Fraud Analytics, Collaborative AI In FinanceAbstract
The detection of financial fraud continues to be an important problem to be solved for all major international banking institutions due to projected annual fraud loss figures over $32B. The primary means to detect fraud effectively, is through the examination of a wide variety of transactions from many different banking organizations. However, due to privacy laws including GDPR and CCPA, as well as competition law related to control of customer data, it is very difficult for banking organizations to share their customers’ transactional data. In this paper we describe a novel federated learning architecture for fraud detection which will enable collaboration among multiple banking institutions to identify fraud without the exchange of customers' private raw transaction data. The proposed method will use a decentralized machine learning architecture where each participant trains its own model using their private dataset and transmits to a centralized coordinating node only encrypted, noisy versions of the model's updates. To ensure the security of the transmitted model updates, we propose to employ secure aggregation protocols, differential privacy mechanisms, and Byzantine-robust aggregation methods to prevent malicious participants from carrying out inference attacks and model poisoning attacks. Experimental results show that our method can achieve a 15 – 30% increase in fraud detection rates relative to stand-alone institutionally-trained models, and are able to find complex cross-institutional patterns indicative of fraudulent behavior. These results also confirm that our method does comply with all relevant regulatory requirements regarding the privacy and sovereignty of customers' data. Thus, this method presents a technical pathway to allow banking consortia to collaborate to counter sophisticated organized crime groups in a manner consistent with applicable data sovereignty, privacy regulations and competitive restrictions.
References
[1] M. A. Salam, D. L. El-Bably, K. M. Fouad, and M. S. E. Elsayed, "Enhancing Fraud Detection in Credit Card Transactions using Optimized Federated Learning Model," Int. J. Adv. Comput. Sci. Appl. (IJACSA), vol. 15, no. 5, May 2024.
[2] H. Dai, "Federated Learning-Based Credit Card Fraud Detection: A Comparative Analysis of Advanced Machine Learning Models," ITM Web Conf., vol. 70, no. 01022, Jan. 2025.
[3] S. Li, E. C. H. Ngai, and T. Voigt, "An Experimental Study of Byzantine-Robust Aggregation Schemes in Federated Learning," IEEE Trans. Big Data, vol. 10, pp. 975-988, Feb. 2023.
[4] Z. Xia and S. C. Saha, "FinGraphFL: Financial Graph-Based Federated Learning for Enhanced Credit Card Fraud Detection," Mathematics, vol. 13, no. 9, p. 1396, Apr. 2025.
[5] A. Awosika et al., "Transparency and Privacy: The Role of Explainable AI and Federated Learning in Financial Fraud Detection," IEEE Access, vol. 12, pp. 33945-33958, Jan. 2024.
[6] M. Talukder et al., "Secure and Transparent Banking: Explainable AI-Driven Federated Learning Model for Financial Fraud Detection," J. Risk Financ. Manag., vol. 18, no. 4, p. 179, Apr. 2024.
[7] A. Zafar and M. Teixeira, "Byzantine-Robust Federated Learning Using Generative Adversarial Networks," arXiv preprint arXiv:2503.20884, Mar. 2025.
[8] M. Aljunaid et al., "Federated Learning for Privacy-Preserving Fraud Detection in Digital Banking: Balancing Algorithmic Performance, Privacy, and Regulatory Compliance," J. Financ. Transform., vol. 60, pp. 45-58, Aug. 2024.
[9] UK Finance, "Economic Crime Plan 2.0 Reports: How federated learning strengthens fraud detection in 2025," London, UK, Mar. 2025.
[10] S. Baghdadi et al., "Hybrid Deep Learning for CCFD: Balancing Response Time and Predictive Capability," Journal of Financial Safety, vol. 4, no. 2, pp. 112-128, Jan. 2024.
[11] S. Li et al., "Cellular Traffic Prediction via Byzantine-Robust Asynchronous Federated Learning," IEEE Trans. Netw. Sci. Eng., Early Access, 2025.
[12] Nilson Report, "Global Fraud Loss Analysis and the Rise of Collaborative Defense," Issue 1256, Dec. 2024.
[13] Ashish Babubhai Sakariya (2021). Relationship Marketing for B2B Success in the Rubber Sector. International Journal of Business Management and Visuals(IJBMV), 4(2), 52-58.
