摘要
传统贝叶斯入侵检测方法未考虑属性和属性权值对检测结果的影响。为此,提出基于特征相似度的贝叶斯网络入侵检测方法。利用相似度对网络连接数据的属性特征进行选择,抽取其关键特征,并降低属性的冗余度,以优化朴素贝叶斯的分类性能。实验结果表明,该方法能降低分类数据的维数,提高分类的准确率。
The traditional Bayesian intrusion detection method can not consider the fact that their different actions have differences between data attributes.This paper uses similarity to select the attribute features of network connecting data,gets the main feature,reduces attribute redundancy to improve the traditional Bayesian classification performance.Experimental results show that this method can reduce the dimension of the classification data,and improve the classification accuracy.
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第21期102-104,共3页
Computer Engineering
基金
天津市教委滨海双百基金资助项目(SB20080053
SB20080055)
教育部科技计划基金资助重点项目(208010)
关键词
特征选择
相似度
贝叶斯分类
入侵检测
feature selection
similarity
Bayesian classification
intrusion detection
