利用GA与SVM对NIDS进行关键特征提取

入侵检测是网络信息安全系统的重要组成部分,而检测特征数量的多少是影响整个入侵检测系统性能的重要因素。介绍了一种减少冗余特征、确定关键特征的方法。这种方法以检测精度为基准,借助遗传算法(GA)寻优,利用支持向量机(SVM)评价,根据统计学原理进行重要性排序。最后按照排序,根据检测精度和误判率变化情况减少冗余,确定关键特征。实验结果理想,并且,与文献1,2]相比,关键特征更少,说明这种方法是科学的,是完全可行的。

Key Feature Selection for NIDS Using GA and SVM

Intrusion detection is a critical component of secure information systems,and the problem of input feature sorting and selection impacted heavily upon the performance of overall intrusion detection system(IDS).This paper address the method of identifying key features in building an IDS.The method uses support vector machine(SVM) to evaluate the detection accuracy in the iterative process of Genetic Algorithm(GA),and sorte the input features by frequency of the feature that has won,then identify the key features by detecting accuracy changed.The result of experiment overlapped approximately the result in reference1,2],moreover,the number of key features is less than those in reference1,2].It proves that the method is scientific and useful.