基于密度聚类分析的入侵检测方法研究

Intrusion Detection Approach Based on Density Clustering Analysis

针对密度聚类DBSCAN算法存在的聚类效果对输入参数敏感的问题,提出了一种基于k-means改进算法确定DBSCAN算法参数的方案来提高聚类质量。将改进k-means算法与DBSCAN算法相结合应用于入侵检测系统,实验结果表明,新方法较好地解决了传统DBSCAN聚类算法中参数选择的敏感问题,相比于李娜等人提出的算法,结合算法使检测率提高了3.32%,误报率降低了1.83%。

With regard to the issue that the clustering results will be affected by the sensitivity of the input parameters in the density clustering DBSCAN algorithm.This paper proposed an improvement plan which determining the DBSCAN algorithm parameters on the basis of k-means algorithm to enhance clustering quality.And the experimental result show that by putting the combination of k-means algorithm and DBSCAN algorithm into the intrusion detection system,this new method can well solved the parameters＇ sensitive issue in the traditional DBSCAN clustering algorithm.This new algorithm can make the detection rate improved by 3.32%,and the false positive rate decreased by 1.83% compared with Li Na＇s algorithm.