Research on Feature Selection Method of Intrusion Detection Based on Deep Belief Network

摘要

Feature selection is one of the important factors that affect the intrusion detection system. Aiming at the problems due to selecting the high feature dimension and the redundancy cause low detection accuracy and high missing rate in the traditional intrusion detection system. In this paper, the deep belief network algorithm is given to select features layer by layer to reduce the feature dimension. As the deep belief network algorithm is an unsupervised learning algorithm, it is more suitable for selecting features from a large number of unlabeled data. Compared with other feature selection algorithm, the experiment shows the deep belief network algorithm is more effective than other algorithm in intrusion detection network.