ENHANCING AI-BASED INTRUSION DETECTION VIA GENERATIVE ADVERSARIAL NETWORKS

Abstract

As Artificial Intelligence (AI) continues to gain importance in the field of cybersecurity [1], it is being used in a variety of areas such as intrusion detection [2], securing industrial networks [3], and improving cryptography [34]. we investigated the impact of using AI for intrusion detection purposes, specifically for detecting DDOS and PortSCAN attacks. Our study indicates that different ML and DL models can achieve great results in terms of metrics. Our findings suggest that different feature selection methods can be applied to also achieve good results even with lower number of features. This research sheds the light also on adversarial attacks and how they affect the performance of the models from cybersecurity and AI perspectives. As the field continues to evolve, further investigation is needed in terms of robustness of the model and the datasets that are being created [4].