入侵检测的CPN改进算法的研究  

A Research into Intrusion Detection Based on Improved CPN Algorithm

作　　者：罗锦光[1] 元昌安[1] 邹鹏[1] 罗锦坤[1] 

机构地区：[1]广西师范学院计算机与信息工程学院,广西南宁530023

出　　处：《广西师范学院学报（自然科学版）》2011年第2期76-80,共5页Journal of Guangxi Teachers Education University(Natural Science Edition)

基　　金：国家自然科学基金(GrantNo.60763012);广西自然科学基金重点项目(2011GXNSFD018025);广西研究生教育创新计划基金项目((2010106030774M02)

摘　　要：结合神经网络方法,对入侵检测技术的聚类分析方法进行了研究和分析,探讨了在传统的对偶传播神经网络(Counter Propagation Networks,CPN)的基础上,引入基因表达式编程(Gene Expression Programming,GEP)对聚类进行优化,提出一种应用于入侵检测的CPN改进算法,该方法融合了GEP算法理念,克服传统的CPN网络算法中输入向量影响Kohonen层的连接权值,导致陷入网络抖动的缺点,提高了网络的收敛时间,得到了更好的网络聚类效果,并从实验中验证了算法在入侵数据分析中的有效性与优越性。Combined with the neural network method,an analysis of clustering in intrusion detection technology is conducted,and the optimization of the clustering by introducing the gene expression programming is discussed based on the counter propagation networks.And an improved CPN algorithm that is applied to the intrusion detection is proposed,which can overcome the problem of the input vector affecting the connection weight value in the Kohonen of the traditional CPN and in the mean time the networks will fall into the tremble.Furthermore,much better clustering result has been obtained,and the convergence time has been increased.Finally,the validity and superiority of the presented algorithm is demonstrated by experiment in intrusion data analysis.

关 键 词：入侵检测 网络入侵 对传网 聚类算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]