基于空间降维和多核支持向量机的网络入侵检测

Network Intrusion Detection Based on Spatial Dimension Reduction and Multi-kernel Support Vector Machine

提出空间降维和多核支持向量机算法进行网络入侵检测;从网络中抓取数据包,通过局部线性嵌入数据降维获得属性降维后的数据样本;通过差异化设置参与局部线性嵌入运算的邻居数,验证适合样本集的邻居数,将多项式核函数、高斯核函数和Sigmoid核函数进行两两组合,分别验证多核支持向量机的网络入侵检测性能。结果表明,在合理设置局部线性嵌入邻居数和核函数组合方式的条件下,基于空间降维和多核支持向量机的网络入侵检出率较高,检测时间较短。

Spatial dimension reduction and multi-kernel support vector machine algorithm was proposed for network intrusion detection.Data packets were fetched from the network,and data samples after attribute dimension reduction were obtained by local linear embedding data dimension reduction.Number of neighbors involved in local linear embedding operation was set by differentiation to verify the number of neighbors in the appropriate sample set.The multinomial kernel function,Gaussian kernel function,and Sigmoid kernel function were combined in pairs to respectively verify network intrusion detection performance of the multi-kernel support vector machine.The results show that the network intrusion detection rate based on spatial dimension reduction and multi-kernel support vector machine is higher and the detection time is shorter when the local linear embedding neighbor number and kernel function combination are set reasonably.