Font Size: a A A

Research Of Intrusion Detection Model Based On Particle Swarm Optimization Algorithm And Support Vector Machine

Posted on:2013-01-01Degree:MasterType:Thesis
Country:ChinaCandidate:F J XiaoFull Text:PDF
GTID:2218330371472036Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
The time of worldwide communication is coming due to the rapid development of computer network technology. The characteristic of network is its openness to everybody. So the protection for privacy and personal information is unavoidable when many users share some resource. Especially, E-commerce and E-government develop rapidly which highlight the importance and urgency of network security. Intrusion Detection Technology is dynamic security technology that can convert static guardian to dynamic defense and make up for the deficiencies of the traditional security. The intrusion approaches have been more complex and diversified. In this case, traditional Intrusion Detection Technology can't satisfy the needs of current network security because of its widespread lower efficiency and the inadequate performance. In order to improve the detection efficiency of Intrusion Detection System, to reduce the rate of error and failure and to reduce the time spent to detect the intrusion, the algorithm of Machine Learning is introduced to Intrusion Detection Field. Appropriate algorithm and Intrusion Detection Model featured by high efficiency and high accuracy have been the consensus in the field.Support Vector Machine and Particle Swarm Optimization are introduced to Intrusion Detection Technology according to the previous analysis of current Intrusion Detection Approaches. At the same time, we research their applications in Intrusion Detection Field. SVM is one of best Machine Learning Approaches which is based on the principle of structural risk minimization. SVM with high precision is mainly applied to Small Sample Learning which isn't sensitive to dimension of data. As we know, the kernel function and parameters will determine the performance of SVM. The artificial specified way is adopted when the kernel function is constructed, which prevents the flexibility and best performance. At the same time, the infinite parameter candidates in the field determined by parameter of SWM increase the time complexity without optimized parameter. When the number of sample increase, the time complexity and space complexity rise sharply, this is one weakness of SVM classifier. In fact, it's the sample data which is from web and isn't related with intrusion action very much that lower the efficiency of classifier.SVM Intrusion Detection Approach encounters problems described above, so a Collaborative network intrusion detection model based on the algorithm Gray Support Vector Machine and Particle Swarm is proposed in the paper. This model takes advantage of gray relational analysis theory to reduce redundant sample in the sample set to reduce the dimension of sample set. We use basic kernel function to construct a mixture of kernels when constructing the kernel function of SVM. At the same time, we optimize the parameter of basic kernel function and the parameter of SVM using particle swarm optimization algorithm. In order to improve the performance of detection model, three intrusion detection agencies based on PSO/SVM do detection work. In this paper, the simulation experiment is carried out on Matlab with KDD cup1999data set. The experiment result shows the proposed model obtains a great capacity of learning and higher quality of classification.
Keywords/Search Tags:Intrusion Detection, Support Vector Machine, Particle Swarm Optimization algorithm, Kernel function
PDF Full Text Request
Related items