基于遗传算法的入侵检测系统特征选择方法研究

Research on Feature Selection Method of Intrusion Detection System Based on Genetic Algorithm

在入侵检测系统中,分类器所选取的特征对系统的性能有很大的影响,大量冗余和不相关特征的存在会降低系统的正确性和实时性,因此如何选取出最优特征组合成为研究的热点问题。在研究当前各种特征选择方法的基础上,提出了一种基于遗传算法的特征组合选择方法。使用遗传算法搜索特征空间,依据Fisher准则计算各种特征组合的分类能力,根据计算结果对特征组合进行选择、交叉、变异,通过多次反复迭代最终选取出最优的特征组合。在实验中分别使用全部特征和选取出的最优特征组合的进行分类验证,最终证明选取出的最优特征组合能够使入侵检测系统在保持高检测率和低误报率的同时具有较高的检测效率,提高了系统的整体性能。

In the intrusion detection system, the features chose by the classifier have a great impact on theperformance of the system. The irrelevant and redundant features can reduce the correctness and the real timeperformance of the system. So, how to select the optimal combination of features has become a hot topic. Onthe basis of researching current methods of feature selection, a method based on genetic algorithm is proposed.The genetic algorithm is used to search the feature space, the classification ability of the feature combinationsare calculated according to the Fisher criterion and do selection, crossover and mutation according to the cal-culation results. This is an iterative process and the optimal combination of features is confirmed at last. Theexperiments are done with both all feathers and the optimal combination of features. The results of the experi-ments show that the intrusion detection system with the optimal combination keeps a high true positive rate anda low false positive rate and at the same time have a high efficiency. The whole performance of the intrusiondetection system is improved.