An Optimized Intrusion Detection System Using PCA and BNN

This paper proposes an optimized Intrusion Detection System (IDS) using Principle Component Analysis (PCA) and Back-propagation Neural Network (BNN). Existing neural network based IDS are mainly suffering from two problems: one is to determine the numbers of hidden layers and regulating weight values to configure its topology. The other is to process the large amounts of audit data. In order to increase detection rates and decrease the processing overheads, we exploit Genetic Algorithm (GA). The operation of GA enables IDS based on combination of PCA and BNN to increase their detection rates and decrease processing overheads. The experimental results on KDD 1999 intrusion detection dataset demonstrate the possibility of our approach.