Enhancing Intrusion Detection in 5G Networks: A Comparative Study of Generative Adversarial and Diffusion Networks

With the improvement of 5G Networks, the threat landscape for cyber-attacks has expanded, challenging intrusion detection due to the intelligent and complex nature of the network. Several solutions are emerging to address this issue with Machine Learning (ML)-based Intrusion Detection Systems (IDS). Even though there are implemented IDS ML models, the next problem comes due to a lack of data to accurately train these models. Due to security and privacy restrictions, most organizations are not allowed to use their own data to train a model for intrusion detection. Therefore, in this research, we investigate the generation of tabular network data including malicious data by using Generative Adversarial Networks (GANs) and Diffusion Networks for intrusion detection in 5G networks. We aim to enhance the anomaly detection capabilities by combining real and synthetic data. We provide extensive comparison on the performance of GAN and Diffusion models and further explore privacy-preserving techniques. Also, we incorporate Explainable AI (XAI) methods to make explanations for the ML models’ predictions on tabular network data.