Federated Learning in the Financial Sector: Fraud Detection and Financial Distress Prediction

This paper considers the method of federated learning – a technique for training machine learning models on distributed data without centralizing it physically. It is shown how the key driver of this technology's adoption in banking was regulatory pressure from GDPR rather than abstract privacy concerns. Detailed consideration is given to the theoretical foundations of federated learning and the FedAvg algorithm, as well as two applied cases – credit card fraud detection (F1 = 77%) and U.S. consumer financial distress prediction using NFCS data (F1 = 42.2%, less than 0.5 p.p. below a centralized baseline). Additionally, gradient inversion attacks, participant data heterogeneity, and legal uncertainty around deployment are examined. The conclusion is that federated learning constitutes a useful yet conditionally applicable tool subject to specific technical and institutional constraints.

1. Kumar A. Redefining Finance: The Influence of Artificial Intelligence (AI) and Machine Learning (ML) / A. Kumar // arXiv [Электронный ресурс]. – URL: https://arxiv.org/abs/2410.15951 (дата обращения: 16.01.2026).

2. Communication-Efficient Learning of Deep Networks from Decentralized Data / B. McMahan, E. Moore, D. Ramage [et al.] // Proceedings of the 20th International Conference on Artificial Intelligence and Statistics, AISTATS 2017, 20–22 April 2017, Fort Lauderdale, FL, USA. – PMLR, 2017. – P. 1273–1282.

3. Sha X. Research on financial fraud algorithm based on federal learning and big data technology / X. Sha // arXiv [Электронный ресурс]. – URL: https://arxiv.org/abs/2405.03992 (дата обращения: 09.02.2026).

4. Carta L. Explainable Federated Learning for U.S. State-Level Financial Distress Modeling / L. Carta, F. Spadea, O. Seneviratne // arXiv [Электронный ресурс]. – URL: https://arxiv.org/abs/2511.08588 (дата обращения: 20.01.2026).

5. Srivastava R.K. Highway Networks / R.K. Srivastava, K. Greff, J. Schmidhuber // arXiv [Электронный ресурс]. – URL: https://arxiv.org/abs/1505.00387 (дата обращения: 13.01.2026).

6. Lundberg S.M. A Unified Approach to Interpreting Model Predictions / S.M. Lundberg, S.-I. Lee // Advances in Neural Information Processing Systems 30: Annual Conference on Neural Information Processing Systems 2017, 04–09 December 2017, Long Beach, CA, USA. – 2017. – P. 4765–4774.