Research On Intrusion Detection With The Improved QPSO To Optimize RBF Network

With the rapid development of computer technology and the wide application of Internet, network intrusion attacks happen frequently, and invasive species emerge in endlessly, making the firewall is more and more difficult to separate to ensure the safety of network. As a result, the enhancement network security, there are many organizations and experts are tend to be more powerful active safety protection strategy, among them, the intrusion detection technology as a kind of effective solutions, has become the focus of research.This paper introduces the intrusion detection technology and it’s development trend. Then we analyzed and summarized the traditional intrusion detection technologies, such as low detection capability to unknown network attack, low real-time detection attack ability defects, high false positive and negative rate. Due to their own characteristics, and artificial neural network is applied to intrusion detection is feasible. In recent years, intrusion detection technique has started using neural network tool, opened up a new intrusion detection approach. Therefore this paper introduces some related theories of Neural Network.This paper focus on the research of RBF neural network(Radia Basis Neural Network) which has been extensively used including the principle, network structure and weight learning process. Introduced in the process of RBF neural network parameter optimization of quantum particle swarm optimization algorithm is a kind of simple calculation, fast convergence speed. But quantum particle swarm optimization algorithm in the late iterations is easy to fall into local convergence precision is not ideal. Then the paper proposes a new improved algorithm. In the early stage of the algorithm can effectively expand the search space, prevent algorithm trapped in local optimum. Double center particles, introduced at the same time, the QPSO algorithm to get the global optimal solution of each dimension of values and double center particles corresponding dimensions respectively replace, update global optimal again, make the algorithm to achieve high search precision and has faster convergence speed. Introduced in the process of RBF neural network parameter optimization of the improved QPSO algorithm can improve the detection of RBF neural network in intrusion detection rate and convergence speed.Lastly, the simulation experiment under MATLAB R2010 a is carried out with the use of partial data in KDDCUP99 database. It makes a comparison of the new improved algorithm that this paper proposed and other methods: QPSO_RBF and IQPSO_RBF algorithm. Through the analysis of the experimental results, it shows that the proposed algorithm achieves the desired consequent.