Deep belief network based intrusion detection techniques: A survey

Abstract

With the recent growth in the number of IoT devices, the amount of personal, sensitive, and important data flowing through the global network have grown rapidly. Additionally, the malicious attempt to access important information or damage the network have also become more complex and advanced. Thus, cybersecurity has become an important issue for the evolution toward future networks that can react and counter such threats. Intrusion detection is an important part of the cybersecurity technology with the goal of monitoring and analyzing network traffic from various resources and detect malicious activities. In recent years, deep learning base deep neural network (DNN) techniques have been utilized as the key solution to detect malicious attacks and among many DNNs, deep belief network (DBN) has been the most influential technique. There have been many attempts to survey wide range of machine learning and deep learning technique based intrusion detection research works, including DBN, but failed to provide a complete review of all the aspects related to the DBN based intrusion detection models. Unlike previous survey papers, we first provide basic concepts on data set, performance metric, and restricted Boltzmann machines, to help understand the basic DBN based intrusion detection model. Finally, a complete review and analysis on the previously published works on DBN based IDS models is provided.