改进的Re-FCBF算法在入侵检测中的应用  

Application of improved Re-FCBF algorithm in intrusion detection

作　　者：沙丽瓦尔.阿里木 努尔布力[1] 吾守尔.斯拉木 Sarwar Alim Nurbol Wushouer Silamu(College of Information Science and Engineering, Xinjiang University, Urumqi 830046, Chin)

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046

出　　处：《计算机工程与设计》2016年第11期2956-2960,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61163052;61303231;61433012);国家自然科学基金联合基金项目(U1435215)

摘　　要：为解决入侵检测领域计算复杂度、时间复杂度高的难题,达到更优秀的入侵检测效果,有效降维,在原有的ReFCBF算法的基础上,提出增强区分特征间互信息的能力,以在改进的Re-ReliefF算法的基础上,实现更佳的入侵检测效果为目标。实验采用DARPA 2000数据集,对数据的41维特征进行选择,采用支持向量机作为分类器,实验结果表明,该改进方法在分类的耗时和误报率略好的情况下,提高了30%的准确率。To solve the problem of high computational complexity and time complexity,to achieve more effective intrusion detection results,an effective dimensionality reduction,on the basis of the original Re-FCBF algorithm,was proposed with the ability of distinguishing characteristics between mutual information,and to achieve more effective dimensionality reduction.On the basis of improvements to Re-ReliefF algorithm,achieving better intrusion detection effects was set as the goal.DARPA 2000 data sets were used to select 41-dimensional feature data for the experiment,and support vector machine was used as the classifier.Experimental results show the improved method proposed increases the classification accuracy by 30% on the premise of slightly reducing time-consumption and the rate of false.

关 键 词：入侵检测 特征选择 Re-FCBF算法 冗余 特征权值 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]