基于距离加权模板约简和属性信息熵的增量SVM入侵检测算法

Incremental SVM Intrusion Detection Algorithm Based on Distance Weighted Template Reduction and Attribute Information Entropyc

为了解决SVM入侵检测方法检测率低、误报率高和检测速度慢等问题,提出了一种基于距离加权模板约简和属性信息熵的增量SVM入侵检测算法。该算法对K近邻样本与待测样本赋予总距离加权权重,对训练样本集进行约简,并以邻界区分割和基于样本属性信息熵对聚类样本中的噪声点和过拟合点进行剔除,以样本分散度来提取可能支持向量机,并基于KKT条件进行增量学习,从而构造最优SVM分类器。实验仿真证明,该算法具有较好的检测率和检测效率,并且误报率低。

In order to solve the problem of the SVM intrusion detection method which has low detection rate,high distorting rate and slow detection speed,a kind of incremental SVM intrusion detection algorithm based on distance weighted template reduction and the attribute information entropy was proposed.In this algorithm,the training sample set reduction is made according to the sample for the samples and the neighbors to the total distance weighted weight,then,the clustering sample point and the noise of the fitting point are taken out through the adjacent to the border area segmentation and based on sample attribute information entropy,and then,using the sample dispersion extracts possible support vector machine,and incremental learning based on KKT conditions is made to construct the optimal SVM classifier.The simulation results show that the algorithm has good detection rate and the detection efficiency,and distorting rate low.