Prediction of Cyberattacks using Machine Learning in the Financial Sector: Technical Review

Torres V.M.; Diaz J.R.; Portugal E.S.; Cornejo J. · ISMSIT 2025 - 9th International Symposium on Multidisciplinary Studies and Innovative Technologies, Proceedings (2025) · DOI: 10.1109/ismsit67332.2025.11267822

Resumen

This systematic literature review (SLR) tests the use of Machine Learning and Deep Learning algorithms to predict cyberattacks and fraud on business portals in the financial industry. The most effective techniques are tested to identify abnormal patterns in digital transactions, achieving accuracies up to 99% and reducing false positives by 31%, as well as current approaches to money laundering, fraud, terrorist financing, and systemic risks on financial platforms. The report identifies the use of hybrid models, deep neural networks, graph neural network (GNN)-based architecture, harvesting techniques, compensation methods, and active learning, such as how to improve the most obvious strategy for threatening predictions. Through the use of transactional data, user profiles, and behavioral metrics, it is possible to create robust and adaptable models ideal for the protection of financial systems. In addition, the increase in the incorporation of emerging technologies such as blockchain and differential privacy in anti-fraud systems is observed. This RSL concludes that applying intelligent models to financial portals improves early detection of cyberattacks; however, it is recommended to continue researching these models in real-time, data-driven environments, as well as promoting the practical implementation of these models in financial institutions.

Instituciones: Universidad Tecnológica del Perú