Federated learning in University IT security: A conceptual framework for privacy-preserving cyber threat detection

As the alarming rate of cyber threats increases in higher education institutions, the challenge of protecting sensitive data while ensuring efficient threat detection becomes more complex. There is a risk of violating data privacy standards such as Family Educational Rights and Privacy Act (FERPA) and General Data Protection Regulation (GDPR) while using traditional cybersecurity methods. Federated Learning (FL) mitigates this by allowing decentralized model training without sharing raw data. This paper proposes a novel conceptual framework for applying FL in university IT security systems. By allowing departments to train local threat detection models without sharing raw data, the framework preserves confidentiality while enabling collaborative learning across institutional silos. This research employs a design science approach outlining the framework’s architecture, key components, privacy-enhancing techniques, and implementation considerations. It also explores the potential benefits such as improved detection accuracy, and regulatory compliance as well as limitations related to system heterogeneity and communication overhead. The study concludes by identifying future directions for pilot implementation. This work contributes to a scalable, adaptable solution for strengthening cybersecurity across the higher education landscape while upholding institutional autonomy and privacy.