基于粗糙集特征约减的网络异常检测方法

Performance Optimization for Network Anomaly Detection Using Rough Set Attribution Reduction

讨论了基于粗糙集特征约简的SVM(支持向量机)异常检测方法,对源自KDD99的实验数据集分别采用SVM和特征约减后SVM进行仿真实验,依据实验结果的比较,说明在网络异常检测中基于特征约减后的SVM和直接采用SVM相比,在保持检测精度不显著降低的同时,前者能够有效的缩短训练时间.

Discussed method of network anomaly detectio n using rough-set-attribution-reduction-based SVM.The testing data set from KDD 99 were experimented by respectively using attribution-reduction-based SVM and original SVM.The experiment results showed that,compared with applying SVM directly,to apply attribution-reduction-based SVM in network anomaly detection can reduce effectively training time without obviously losing detection accuracy.