基于粗糙集分类的网络入侵检测

提出使用粗糙集分类(RSC)算法进行智能化的网络入侵检测.该方法可以在生成检测规则之前完成特征排序,且不需要多次重复迭代计算,提高了入侵检测系统的效率;同时,生成的检测规则是"if-then"格式的产生式,易于解释.仿真实验表明,RSC对Probe和DoS攻击具有比支持向量机(SVM)略好的高检测率,但是训练时间比SVM更长,采用混杂遗传算法求解粗糙集约简可进一步减少RSC的训练时间.

An Intrusion Detection Mechanism Using Rough Set Classification (RSC)

This paper presented an intelligent intrusion detection mechanism using Rough Set Classification (RSC). Given some attacks, the intrusion detection mechanism using RSC can get both explainable detection rules and high detection rate. Furthermore, RSC can also accomplish fast feature ranking for the intrusion detection system. The performance of RSC was compared with that of Support Vector Machine (SVM), which is a classical intrusion detection algorithm. Experiments were carried out based on a set of benchmark DARPA data. It is observed that both RSC and SVM obtain high accurate results (with accuracy greater than 99% on testing set) for Probe and DoS attacks. However, RSC has a higher training time in comparison with the SVM. To solve this problem, an improved hybrid genetic algorithm was proposed here to expedite the convergence speed of computing the rough set reducts and decrease the training time of RSC.