| 基于特征筛选的加权SVM入侵检测系统 |
| |
| 引用本文： | 王耀普,潘晓中.基于特征筛选的加权SVM入侵检测系统[J].计算机安全,2014(1):6-10. |
| |
| 作者姓名： | 王耀普 潘晓中 |
| |
| 作者单位： | [1]武警工程大学电子技术系网络与信息安全武警部队重点实验室,陕西西安710086 [2]武警工程大学电子技术系网络与信息安全研究所,陕西西安710086 |
| |
| 基金项目： | 武警工程大学基础研究基金资助计划（WJY201125） |
| |
| 摘 要： | 为了克服传统支持向量机中弱相关特征对分类器的分类效果的干扰及二分类SVM入侵检测算法缺乏高效率和低准确率的问题，因此需要优化SVM算法、以保证IDS能够检测出存在的入侵行为。分析了当前主流SVM算法及其发展，通过采用灰色斜率关联分析方法筛选主特征，再用增益比率法对特征进行加权，减少弱相关特征对分类的影响．提出了改进的支持向量机算法。实验证明，本文异常检测系统在检测准确率、检测精度上都有优良的性能。
|
| 关 键 词： | 支持向量机 入侵检测系统 增益比率法 灰色关联分析 |
English Title IDS based on Feature-sifting and Weighted SVM |
| |
| WANG Yao-pu,PAN Xiao-zhong.English Title IDS based on Feature-sifting and Weighted SVM[J].Network & Computer Security,2014(1):6-10. |
| |
| Authors: | WANG Yao-pu PAN Xiao-zhong |
| |
| Affiliation: | 1. Key Laboratory of Network and InformatlOn Securty Under the Armed Police Force, of Electronic Technology, Engineerling University of the Chlnese Armed Police Force, Xi＇an 710086, China 2. Computer College, Northwestern Polytechnical University. Xi＇ an 710072, China) |
| |
| Abstract: | To conquer the problem that the traditional SVM has weak correlation feature which will affect the classification effect of the classification machines, and ＂the binary classification SVM algorithm are low in efficiency and accuracy, the SVM algorithm needs be optimized to ensure the detect of the intrusion behaviors by the IDS. In this paper, the current main SVM algorithm and their development are reviewed. A improved SVM algorithm is proposed, first by sifting the main features using grey correlation analysis, then the features are powered by using gain ratio, which reduces the affection of the weak correlation. The experiment results show that our IDS has a goocl performance both in accuracy and precision. |
| |
| Keywords: | SVM IDS gain ratio grey correlation analysis |
| 本文献已被 维普 等数据库收录！ |
