This paper proposes a Hybrid Kernel-based Support Vector Machine (HKSVM) learning technique for Intrusion Detection System. To improve the performance of SVM, a hybrid approach is proposed as a kernel. The proposed hybrid kernel suggested in this paper is a combination of Polynomial kernel and Gaussian kernel approaches. This model is implemented and evaluated experimentally on a benchmark network intrusion detection dataset known as Kyoto 2006+ dataset and compared with the traditional SVM method. Results are proven that the accuracy of the proposed approach is improved by 9.06% when compared with Gaussian kernel-based SVM model and about 6.87% when compared with Polynomial kernel-based SVM model with a very less false positive rate.
Volume 11 | Issue 8