Efficient Intrusion Detection Based on Multiple Neural Network Classifiers with Improved Genetic Algorithm

Abstract

The security of computer network is one of the most important issues for all the users. Intrusion may lead to terrible disaster for network users. Therefore, it is imperative to detect the network attacks to protect the information security. The intrusion patter identification is a hot topic in this research area. Using artificial neural networks (ANN) to provide intelligent intrusion recognition has been received a lot of attentions. However, the intrusion detection rate is often affected by the structure parameters of the ANN. Improper ANN model design may result in a low detection precision. To overcome these problems, a new network intrusion detection approach based on improved genetic algorithm (GA) and multi-ANN classifiers is proposed in this paper. The improved GA used energy entropy to select individuals to optimize the training procedure of the BPNN, RBF, PNN and Fuzzy-NN. Then, the satisfactory ANN models with proper structure parameters were attained. In addition, to alleviate the complexity of the input vector, the principal component analysis (PCA) has been employed to eliminate redundant features of the original intrusion data. The efficiency of the proposed method was evaluated with the practical data, and the experiment results show that the proposed approach offers a good intrusion detection rate, and performs better than the standard GA-ANN method.