AI-Powered Intrusion Detection and Privacy Preservation in 6G Networks

The arrival of 6 G networks offers unprecedented connectivity, ultra-low latency and large-scale device deployments, making them an exceptional target for sophisticated cyber threats. Existing intrusion detection methods cannot meet 6G's dynamic and heterogeneous ecosystem; therefore, there is a serious need to implement an Artificial Intelligence (AI) - based security architecture. In this work, an AI-based Intrusion Detection System (AI-IDS) that relies on a decentralized learning approach and privacy preservation to deliver real-time threat detection while preserving users' privacy, has been proposed. The proposed system takes advantage of deep reinforcement learning, while also using adversarial robust models, to expose dynamic cyber-attacks, such as zero-day attacks and adversarial manipulations. Using a federated learning framework that respects privacy, model training is shared in a decentralized fashion over distributed edge nodes while preventing data exposure attacks with high accuracy. The routine of the framework is demonstrated against the existing intrusion detection frameworks on real-world 6G traffic datasets, achieving $\mathbf{2 0 \%}$ improved result detection rate and 35% reduced false positives. A comprehensive security analysis is also provided to evaluate resilience to adversarial attacks, model poisoning and inference attacks. The results illustrate the key role of AI-based security paradigms in enhancing 6G networks to provide scalable, robust and privacy-preserving cybersecurity.