一种半监督网络入侵检测系统SSIDS-CV  

A Semi-supervised Network Intrusion Detection System SSIDS-CV

作　　者：刘宁[1] 

机构地区：[1]商洛学院经济与管理学院,商洛726000

出　　处：《计算机与数字工程》2015年第4期648-651,共4页Computer & Digital Engineering

基　　金：陕西省教育厅科研计划项目(编号:12JK0748)资助

摘　　要：为了提高网络入侵检测的性能,提出一种基于半监督学习的网络入侵检测系统SSIDS-CV。系统由网络嗅探器、训练集生成器和半监督分类器三部分组成。通过对无标记入侵数据进行伪标记,将伪标记后的样本加入到有标记数据集中,参与交叉验证,选取能使分类器误差最小的标记作为最终的标记,扩充有标记数据数目,训练入侵检测分类器。使用KDD Cup 99数据集模拟半监督入侵检测过程,实验结果表明SSIDS-CV能有效地挖掘未标记入侵数据信息,具有较高的入侵检测率。In order to improve the performance of network intrusion, an intrusion detection modal SSIDS-CV was pro- posed. The model was composed of network sniffer, generator for training set and classifier. Unlabeled intrusion data was labeled and added to the labeled data set to participate in cross validation. The labels making classifier error minimum were selected as the final the label to mark and train the classifier. Semi-supervised classification experiment was simulated using KDD Cup 99 data sets. The result showed that SSIDS-CV can effectively dig the unlabeled samples information of intrusion data and has a higher intrusion classification rate.

关 键 词：半监督学习 入侵检测 交叉验证 KDD CUP 99 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]