Network security using machine learning

Abstract

Network Security is one of the major challenges for any organization. With the increase in cyber attacks, Network Security is also a point of concern for home users. Network Intrusion Detection Systems (NIDS) are widely used for securing a network against attacks. Traditional pproaches require regular maintenance like signature updates. Using machine learning a system can be created that can learn from its feedback and improve. This thesis introduces a Generative Adversarial Networks based Network Intrusion Detection System. A Generative Adversarial Network (GAN) is a generative algorithm which learns the data distribution it is trained on and can generate similar examples to the dataset. This GAN, which has learned the properties of benign network traffic, can then be used to identify anomalies in connections. Using this we can create an anomaly detection system which can detect network attacks.