Machine Learning for Intrusion Detection Systems

Abstract

The growth of the internet and network-based services bring to us a lot of new opportunities but also pose many new security threats. The intrusion detection system (IDS) has been studied and developed over the years to cope with external attacks from the internet. The task of an IDS is to classify and stop the malicious traffic from outside to enter the computer system. In recent years, machine learning-based IDS has attracted a lot of attention from the industry and academia. The IDS based on state-of-the-art machine learning algorithms usually achieves a very high predictive performance than traditional approaches. On the other hand, several open datasets have been introduced for the researchers to evaluate and compare their algorithms. This chapter reviews the classification techniques used in IDS, mostly the machine learning algorithms and the published datasets. The authors discuss the achievements and some open problems and suggest a few research directions in the future.