Privacy-Preserving Cyber Threat Information Sharing Leveraging FL-Based Intrusion Detection in the Financial Sector

Abstract

This paper presents an architectural proposal for enhancing anomaly detection in the CyberSec4Europe project use case Open Banking. It proposes a trusted privacy-preserving ecosystem of threat intelligence platforms, based on MISP, to automatically exchange and process cyber threat information in an auditable and privacy-preserving manner. Additionally, a Federated Learning scheme is deployed to share machine learning models trained on a synthetic fraud transactions dataset, and the impact of data anonymization on model accuracy is measured and analyzed. This proposal provides a valuable contribution to the development of robust and efficient threat detection systems to enhance the resilience of organizations in the financial sector.