PRIVACY-PRESERVING FEDERATED LEARNING FOR CYBERSECURITY IN DISTRIBUTED CLOUD NETWORKS

Abstract

Distributed cloud networks pose many security and privacy issues coupled by complex cyber threats, that need to be handled, not only in the realm of data security and privacy, but also to collaborate on threat detection. This research paper looks at the benefits and hurdles signifying Privacy-Preserving Federated Learning (PPFL) in the area of Cybersecurity through distributed machine learning systems. It evaluates differential privacy, secure aggregation and homomorphic encryption possibilities for protecting the sensitive information of a network and offers collaborative intrusion detection. The proposed framework evaluates cyber security issues based on publicly available cyber security datasets incorporating accuracy, precision, recall, F1 score, communication overhead, computation efficiency and adversarial resilience. The results of the analysis indicate the potential of PPFL to provide both privacy protection and cybersecurity performance. Scalable federated architectures are recommended in the study where secure, reliable and privacy protected distributed cloud architectures are developed.