摘要
入侵检测系统(IDS)在处理高维数据时具有计算量大、占用计算机资源较多、训练和预测时间较长等缺点,这就需要对数据在确保有用信息不丢失的前提下进行约简。递归支持向量机(R-SVM)根据各个特征在svm分类器中的贡献大小从分类结果中提取使分类器性能最好的特征,以实现维数约简的目的。将R-SVM理论引入入侵检测系统中,提出了一种基于R-SVM入侵检测方法。通过对KDDCUP99数据集中10Percent数据子集的测试实验结果表明,与用粗糙集做特征提取及传统的几种分类算法相比,用R-SVM做特征提取并结合SVM分类算法用于IDS中的性能较好;与使用全部特征构建的支持向量分类器相比,前者能在保障较好的分类精度的同时,降低训练和预测时间。
The intrusion detection system （II）S） has some defects in dealing with high dimensional problem, such as over loaded, occupying too much resource, long training and forecasting time. therefore, the simplify of practical information becomes such a necessity. Recursive support vector machine （R-SVM） extracts main features from the result of SVM classifier according to the different contribution in it. The R-SVM theory is applied into IDS and an intrusion detection method based on R-SVM is pro- posed. Acorrding to the 10 percent data subset of KDDCUP99 , comparing with Rough Set and various kinds of traditional clas- sification algorithms , The experimental results show that R-SVM is better than rough set, and the training and predicting time of SVM based on R-SVM is shorter than SVM based on full features, without reducing accuracy obviously.
出处
《计算机工程与设计》
CSCD
北大核心
2012年第10期3777-3782,共6页
Computer Engineering and Design
基金
陕西省自然科学基金项目(2009JM7007)
陕西省教育厅专项科研计划基金项目(08JK354)
关键词
入侵检测系统
高维数据
约简
特征提取
递归支持向量机
支持向量机
Key words： intrusion detection system
high dimensional problem
reduction
feature extraction
recursive support vector ma-chine
support vector machine
