一种代价敏感的入侵特征选择算法

入侵特征选择可以有效地提高入侵检测系统的正确率及检测速度。在分析了现有的入侵特征选择算法只考虑特征个数和特征子集分类能力的缺陷后,提出一种代价敏感的入侵特征选择算法。该算法基于免疫克隆选择与粗糙集,在分析特征子集的执行代价的基础上,提出以执行代价和分类能力作为亲和度函数指导的免疫克隆进化算法。在KDDCUP99数据集上进行了测试,测试结果证明该算法可以有效地提高分类准确性,降低入侵检测系统的执行代价。

Intrusion Feature Selection Algorithm with Cost Sensitive

Intrusion feature selection can improve the correctness and detection rate of the intrusion detection system effectively.The existing intrusion feature selection algorithm has the defect that just considers the number of feature and the sorting ability of feature subset,so this paper put forward a cost sensitive intrusion feature selection algorithm.Based on immune clone selection,rough set,and analyzing the execution cost of feature subset,it put forward taking execution cost and sorting ability as affinity function to conduct immune clone algorithm to evolve.Test on the data of KDD CUP 99 was made,and the result proved that this algorithm can improve the correctness and reduce the execution cost of the intrusion detection system.