基于数据挖掘的入侵特征选择与构造的新方法

Research on Data Mining-based Intrusion Feature Selection and Construction

入侵检测问题实际上是一个分类问题,特征选择的好坏直接决定了分类模型的性能。针对计算机安全问题是事后于计算机系统设计、没有标准的审计机制和专门的数据格式用于入侵检测分析用途的现状,讨论了通过扩展数据挖掘基本算法来对分析数据源进行特征选择,同时比较挖掘出来的正常模式和异常模式,构造新的特征,以加强入侵检测准确率和实时性。

Intrusion detection is a classification issue, performance of classification model is determined by feature selection. Computer security dont be considered when designing computer system, there is no standard auditing mechanism and specific data format for intrusion detection analyst. This paper discusses feature selection method through extending basic data mining algorithm analyzing data source, constructing new features to increase accurate intrusion detection rate through comparing mined normal patterns and anomaly patterns.