摘要
数据降维是基于模式识别方法的入侵检测系统需要解决的一个问题。由于主成份分析方法具有两个我们期望的特性,一是不同的主成份之间互不相关,二是每个主成份都是所有原始特征的线性组合,所以将主成份分析应用到入侵检测系统的特征提取中。首先我们使用ReliefF算法去除原始特征中与分类无关的特征,然后再进行主成份分析。在实际的数据集KDDCUP’99上进行的实验结果表明提出的方法是有效及实用的。
The first problem which should to be solved in intrusion detection system based on pattern recognition method is reducing data dimentions. Because principal component analysis has two attributes we expected, one is that various principal component is not relevant, and the other is that each principal component is linear combination of all original features, the authors apply the principal component analysis to extracting features from intrusion detection system. Firstly, we use ReliefF to get rid of features that are irrelevant with classification from original features, then employ principal component analysis. The experimental results on real KDD CUP＇99 dataset show that the proposed method is effective and practicable.
出处
《微电子学与计算机》
CSCD
北大核心
2006年第7期90-92,95,共4页
Microelectronics & Computer
基金
国家自然科学基金项目(60475007)
教育部跨世纪人才基金项目
关键词
入侵检测系统
特征提取
主成份分析
冗余特征
Intrusion detection system, Feature extraction, Principal component analysis, Redundancy feature
