Enhancing the Cyber Security Intrusion Detection based on Generative Adversarial Network

Abstract

We will introduce a survey of profound learning methods, data settings and a comparative analysis for the identification of cyber security intrusions. In particular, we study intrusion detection systems using profound learning approaches. Deep learning (DL) approaches: if the research is based on DL for intrusion systems. DL approaches, if machine learning for intrusion detection systems (IDS) is included in the report. Evaluation of approaches to deep learning: it shows whether DL approaches for IDS are evaluated. We research success in two classification categories under two new real-world data sets for each model, namely, CSE-CIC-IDS2018 and the Bot-IoT dataset in order to deeply learn each model. Furthermore, we use the main metrics of success to evaluate the effectiveness of several methods: false alarm rate, detection rate and Accuracy. This shows that the Generative Adversarial Network (GAN) produces much better outcomes than other state-of-the-art approaches.