摘要
在利用聚类进行入侵检测的方法中,有效地进行聚类是关键。为了对未标识数据进行聚类,提出了一种新的无监督入侵检测方法。该方法克服了聚类算法中对数据输入顺序敏感和需要预设聚类数目的缺点,减少了所需参数个数。通过初始聚类簇的建立和混合遗传算法对初始聚类进行优化组合两阶段的方法来实现聚类,克服了初始聚类对结果的影响,提高了聚类质量,并进行检测入侵。实验结果表明该方法有较好的检测率和误检率。
Among the methods of intrusion detection with clustering, how to make clustering effectively is a critical problem. A new unsupervised clustering method was proposed in this paper. The method can overcome the shortcoming of sensitivity to the order of input dates and necessity to know the cluster number before it works, and it also can reduce the number of parameters＇. By the two stage of creating initial clusters and optimation of the flint stage with hybrid genetic clustering method, it can overcome the influence of initial clusters on results and improve the clustering quality, and then this method was used to detect intrusion. Experimental results demonstrate that this method has better detection rate and false positive rate.
出处
《计算机应用》
CSCD
北大核心
2008年第2期409-411,共3页
journal of Computer Applications
基金
湖南省自然科学基金资助项目(05JJ40133)
关键词
入侵检测
聚类
混合遗传算法
intrusion detection
clustering
hybrid genetic algorithm
