Application of an Improved SVM MultiClass Classification to Intrusion Detection

Intrusion detection system as the key technology of network security becomes research hot spot of the current network security, while precision and generalization performance is the key point of intrusion detection algorithm. According to binary tree method and the characteristics of sphere structured support vector machine, an improved SVM multiclass classification algorithm is proposed to intrusion detection. This algorithm uses similarity functions as weight value and selects two kinds of sample similarity minimum to structure twoclass classifier; to bottomup structure kinds of twoclass classifier of sphere structured SVM. Finally it is applied to intrusion detection. The KDD CUP 1999 intrusion detection data used to simulate experiments. Experimental results show that the algorithm effectively improved the detection accuracy and generalization performance.