Network intrusion detection based on heterogeneous distance

Separation measure plays a decisive role in the multi-class support vector machine. Combine with Heterogeneity of network connection data, this paper propose improved HTSVM based on heterogeneous distance. Firstly, define the heterogeneous distance; then compute the separation measure according to the heterogeneous distance between each two classes; then compute the sample that should belong to class according to separation measure. Finely, this method is applied to the network intrusion detection. And simulate with KDDCUP1999 dataset. Experiment result show the method can accurately measure the similarity between heterogeneous data and improve the accurate rate of classification accuracy.