粒子群优化的半监督入侵检测算法  

PSO-based Semi-supervised Intrusion Detection Algorithm

作　　者：贾志伟[1] 关忠仁[2] 赵建芳[3] 

机构地区：[1]成都信息工程学院计算机学院,四川成都610225 [2]成都信息工程学院信息中心,四川成都610225 [3]江南大学物联网工程学院,江苏无锡214000

出　　处：《成都信息工程学院学报》2012年第3期233-238,共6页Journal of Chengdu University of Information Technology

基　　金：四川省教育厅2010年科研资助项目(10ZA163)

摘　　要：为解决无监督入侵检测算法检测率低,有监督的入侵检测不能有效的检测未知攻击的问题,提出了一种粒子群优化的半监督入侵检测算法,算法对少量的约束信息进行基于密度的扩展获得潜在约束得到聚类模型,以此指导未标记数据聚类,对仍没有确定类别的未标示数据使用粒子群优化的K均值算法进行聚类实现对异常的检测。改进的算法检测率达到83.7%,误报率减少至3.13%,总体效果优于无监督和有监督学习的入侵检测算法。View of unsupervised intrusion detection algorithm has low detection rate, supervised intrusion detection can not detect unknown attacker, this paper proposed a PSO-based semi-supervised Intrusion detection algorithm, it will expand a small amount of constraint information with density expansion method to obtain the clustering model, which used to guide the unlabeled data clustering, at last the unmarked data which was not determined its category, which using particle swarm optimization k-means algorithm to cluster to achieve the detection of abnormal. Algorithm improved the detection rate to 83.7% and the false positive rate decreased to 3.13 %, the overall effect is superior to intrusion detection based on unsupervised and supervised learning.

关 键 词：计算机应用技术 数据挖掘 半监督聚类 入侵检测 约束扩展 粒子群优化算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]