基于无指导离群点检测的网络入侵检测技术  

Network Intrusion Detection Technology Based Unsupervised Outlier Detection

作　　者：黄俊[1] 韩玲莉[1] 陈光平[1] 

机构地区：[1]中国计量学院计算机科学系,浙江杭州310018

出　　处：《小型微型计算机系统》2007年第11期2007-2009,共3页Journal of Chinese Computer Systems

基　　金：浙江省06自然科学基金项目(Y109456)资助.

摘　　要：讨论了基于无指导离群点检测的网络入侵检测技术及实现框架.技术方法首先在网络数据包上通过改进的随机森林算法建立了网络服务模型,然后通过确定网络服务模型上的离群点实现网络入侵检测.还通过在KDD'99数据集上对所提出的技术实现入侵检测的实验及结果进行了讨论并与其他无指导异常检测方法进行了比较.In this paper,the framework of network intrusion detection technology Based unsupervised outlier detection is discussed. In the framework, patterns of network services are built by random forests algorithm over traffic data. Intrusions are detected by determining outliers related to the built patterns. The modification is presented on the outlier detection algorithm of random forests. The experimental results over the KDD＇99 dataset also is reported . The results show that the proposed approach is comparable to previously reported unsupervised anomaly detection approaches evaluated over the KDD＇99 dataset.

关 键 词：网络入侵检测 随机森林算法 无指导离群点检测 网络服务模型 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]