Study On The Mothod For Intrusion Detection Based On Improved Pso Rbf Neural Network

With the rapid development of Internet and computer,Networking applications also will be increased,but,the threat of intrusion for the system and network information has been more and more serious,coupled with today’s attackers’knowledge increasingly mature、 hacking technology level continuously improveing and the tools used by the attacker is also becoming more diverse and complex,finally,resulting in the traditional security methods have been unable to keep up with the pace of network security needs, so that the system and network information security issues become increasingly prominent.Intrusion detection system gradually evolved into one of a proactive network security and defense technology, which effectively compensate for the lack of traditional security technologies. Especially in recent years, IDS (Intrusion-Detection System, IDS) research has made great progress. But, under the Network attack new and changing circumstances the traditional methods which through the characteristics to detect intrusion of the shortcomings gradually exposed, especially when constantly updated in the structure of the network upgrade and unknown forms of attack are becoming increasingly complex and volatile,the lack of necessary scalability and adaptability,artificial neural network (based Artificial Neural network, referred to as ANN) to meet these two requirements.In recent yeas, RBF neural networks (Radial Basis Function Neural Network, referred to as the RBFNN) has been the focus on an artificial neural network at the same time,in many areas,such as classification、 system identification、 function approximation signal processing and so on,has been successfully applied. But in the design of the RBF neural network, RBFNN has three very important parameters (weights w, the center c and the base width σ), the impropriety of their initial value, likely to cause the RBF neural network into a local optimum, convergence slow and precision and so on, But the particle swarm optimization (particle of Swarm optimization, referred to as PSO) algorithm has strong global optimization what can solve these issues. In order to further improve the performance of the RBF neural network and detection rate of the intrusion detection system, the main content of this study are as follows:1. A multi-strategy to improve the PSO algorithm-the IPSO was Proposed,and the algorithm is on the basis of PSO algorithm, Firstly,useing the adaptive inertia weight strategy to adjust the inertia weight to solve the problem which the optimization capabilities of the PSO algorithm for nonlinear problems is low;Secondly, according to the group fitness variance to determine the PSO algorithm which is the phenomenon of "premature convergence"or not;Finally, introducing the mutation operator, if the occurrence of premature convergence,we will do the extreme value disturbance for the optimal solution of the groups particles.By simulating to verify the performance of the IPSO algorithm.2. To further enhance the convergence speed and accuracy of the RBFNN learning algorithm,An algorithm based on improve the PSO algorithm and RBF neural network is proposed, After using the subtractive clustering algorithm to determine the hidden nodes of RBFNN number,this paper combines the IPSO algorithmglobal and the gradient descent method to global optimization and local optimization the parameters of the RBFNN,and uses the Hermit polynomial simulation to verify the performance of the algorithm.3. In order to improve the detection rate of intrusion detection system, an intrusion detection model based on improved PSO algorithm and RBF neural network was proposed. In order to verify the performance of the model, After normalizing and processing, this paper use the data sets as the neural network input data, use the Matlab software to simulation the model.The experimental results show that this method improves the detection rate of intrusion detection systems, and has good efficiency and scalability.