摘要: 入侵检测系统所得原始特征通常是高维度的,这些高维度特征带来了较大的计算开销。针对该问题,采用核典型相关分析方法进行原始特征的二次提取,得到简约而重要的二次特征。在该二次特征的基础上运用二叉树多分类支持向量机法判别待测网络状态所属类别。仿真实验证明,该算法在不显著损失检测准确度的情况下可提升系统实时性,从而达到优化目标。
关键词:
入侵检测系统,
核典型相关分析,
二次特征,
二叉树支持向量机
Abstract: In Intrusion Detection System(IDS), the original features normally lead to considerable computational complexity because of their high dimensions. In this paper, reduced and important further features are obtained by introducing Kernel Canonical Correlation Analysis(KCCA), and the binary tree based multi-class classification SVM is used to complete the classification task by these further features. This algorithm is devoted to increase the real-time performance as much as possible under the condition of not clearly losing classification accuracy. Simulation experiment confirms the above advantages.
Key words:
Intrusion Detection System(IDS),
Kernel Canonical Correlation Analysis(KCCA),
further features,
binary tree based SVM
中图分类号:
钱鹏江;王士同;徐 华;颜惠琴. 基于KCCA优化的网络入侵检测算法[J]. 计算机工程, 2009, 35(23): 118-119.
QIAN Peng-jiang; WANG Shi-tong; XU Hua; YAN Hui-qin. Network Intrusion Detection Algorithm Based on KCCA Optimization[J]. Computer Engineering, 2009, 35(23): 118-119.