改进贝叶斯分类算法在入侵检测中的研究  

Research of Improved Bayesian Classification Arithmetic in Intrusion Detection

作　　者：张铮[1] 高志森[1] 李俊[1] 

机构地区：[1]南京航空航天大学信息科学与技术学院,江苏南京210016

出　　处：《计算机技术与发展》2007年第1期174-176,共3页Computer Technology and Development

基　　金：国防科工委国防基础科研项目(S0500B003)

摘　　要：把朴素贝叶斯分类算法引入到入侵检测中,可以简单方便地区别出入侵事件。但是由于该算法在学习中存在一定的不足和缺陷,主要是属性值之间要求相互条件独立和训练集数据不完备这两个缺陷,导致了它的检测效果并不是很理想。文中针对该算法这两个最主要的缺陷,提出增量学习概念,引入损失幅度参数,改进和完善朴素贝叶斯分类算法。并对改进后的新学习策略进行了分析和研究,给出了其基本实现思想和算法描述,并指出它实现的可能性。Native Bayesian Classification Arithmetic distinguishes intrusion incidents from all incidents conveniently in intrusion detection. But this arithmetic has some lack and limitation, two important factors of which are condition independency among property values and non - maturity training - data collection, so its intrusion effect is not good. Then in allusion to these two important factors, a concept of incremental learning and a loss extent parameter are put forward in this paper, and Native Bayesian Classification Arithmetic is also perfected, Furthermore, the paper analyses the new learning strategy, and introduces the basic realizing thought about new arithmetic and a description about it, and points out its realizing possibility.

关 键 词：贝叶斯分类 增量学习 损失幅度参数 类别标签 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]