入侵检测中的混合特征选择算法研究  

Research on Hybrid Feature Selection Method in Intrusion Detection

作　　者：黎银环[1] 钟艳花[1] 

机构地区：[1]广东省江门职业技术学院,广东江门529000

出　　处：《微电子学与计算机》2012年第9期51-54,共4页Microelectronics & Computer

基　　金：国家自然科学基金项目(60903025)

摘　　要：在开放式网络中,高维混合特征的冗余或不相容属性会降低网络入侵检测的效率.为提高入侵检测系统的响应性能,提出一种混合特征选择方法,利用粗糙集形式化描述入侵检测的特征选择,采用信息熵和平均权重分别定义数值型和字符型特征的重要度.算法产生降序特征序列,采用K-means聚类算法评估出优化特征子集.在KDD CUP99数据集上的仿真实验表明,算法有效选择特征子集并缩短了检测时间.In an open network, redundant or incompatible attributes of high-dimensional mixed features reduce the efficiency of network intrusion detection. In order to improve the response performance of intrusion detection system, this paper proposes a hybrid feature selection method. Rough set theory is used to description for intrusion detection feature selection. Information entropy and average weight are used to define importance of numeric features and character features. After generating a descending feature sequence, K-means clustering algorithm is used to evaluate the optimal feature subset. Simulation experiment is done in KDDCUP99. It shows that the method is effective to select feature subset and shorten the detection time.

关 键 词：粗糙集 特征选择 信息熵 平均权重 重要度 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]