Efficient intrusion detection system based on support vector machines using optimized kernel function

An efficient intrusion detection system requires fast processing and optimized performance. Architectural complexity of the classifier increases by the processing of the raw features in the datasets which causes heavy load and needs proper transformation and representation. PCA is a traditional approach for dimension reduction by finding linear combinations of original features into lesser number. Support vector machine performs well with different kernel functions that classifies in higher dimensional at optimized parameters. The performance of these kernels can be examined by using variant feature subsets at respective parametric values. In this paper SVM based intrusion detection is proposed by using PCA transformed features with different kernel functions. This results in optimal kernel of SVM for feature subset with fewer false alarms and increased detection rate.