Intrusion Detection Based on a Fusion Classifier and SVM
Junguo Liao · Jisuanji fangzhen · 2008
In order to gain the result of identifying a good detection mechanism in intrusion detection,a fusion classifier,which is popularly known as the ensemble approach,is designed to build the intrusion detection system.This fusion classifier contains three classifiers: the first is a SVM based on attribute selection,the second is a SVM based on sample reduction,and the last is a standard SVM.The ensemble approach includes three steps.The first step is data processing.The second step is to carefully construct the different connectional models to achieve the best generalization performance for classifiers.The last step is to form the final decision.The remarkable highlight is choosing the optimal weights strategy.In the performance of the intrusion detection system,a weight value optimization strategy is performed based on the accuracy of a given data class actually classified by each classifier respectively.In fact,the experiments show that Intrusion Detection performances can be improved by combining an ensemble of SVM classifiers.