Deep Learning algorithm for Intrusion Detection System: a survey

Rapid advances in the Internet and communications have greatly increased the size of the network and networks related data. As a result, many new attacks are emerging, challenging network security to accurately identify intrusions. In addition, the presence of hackers with the aim of carrying out various attacks within the network cannot be ignored. Intrusion detection systems are an important protection tool for the network. Intrusion detection systems are classifiers that receive input records and predict the class of types of attacks. In network attacks, there are various deep learning algorithms that have been proposed for intrusion detection systems. Over the past decades, researchers have used a variety of deep learning algorithms to classify and detect malicious traffic from normal traffic without prior knowledge of attack pattern. This article provides an overview of deep learning algorithms that use for intrusion detection systems. A separate section is dedicated to presenting the datasets used in intrusion detection systems in particular, the two main datasets, KDDCup99 and NSL-KDD. Evaluation criteria and implementation tools in intrusion detection systems are also examined.