Research On Multi-Classification SVM Intrusion Detection Based On CMABC Parameter Optimization

With the development of the Internet, people pay more attention on information security, intrusion detection technology has become an important information security technology, it has become a new focus in recent years. However, most existing intrusion detection system exist the problem of low speed, low detection rate. Support Vector Machine can overcome the shortcomings of traditional classification algorithm, it is very suitable for intrusion detection system.For SVM parameters choice determines its learning and generalization ability, its performance largely depends on the parameters penalty factor C and kernel parameter g. This paper analyzes the relationship between parameters and their influence on the performance of SVM, introduce the cross mutation operator in artificial bee colony algorithm, propose the algorithm of optimize SVM based on cross mutation artificial bee colony. Artificial Bee Colony (ABC) algorithm is a new swarm intelligence optimization algorithm, it simulate the natural process of honey bees, the bees complete each phase of the task depending on the division of labor, by collecting and sharing food sources information to fins the optimal solution, it has a few parameters, and simple, the crossover operator introduced effectively reduce the risk of falling into local optimal solution.Traditional SVM is to solve the two problems, can't meet the needs of multi-classification intrusion detection, this paper analyzes the common multi-classification SVM methods, for the shortage of binary tree SVM algorithm proposed class separability measure concepts, build an effective binary tree structure, make the multi-class problem into a series of binary classification problem, and put forward a layered structure according to the data characteristics of the intrusion detection, effectively improve the detection of multi-classification SVM intrusion detection.