期刊名称：International Journal of Security and Its Applications
印刷版ISSN：1738-9976
出版年度：2015
卷号：9
期号：11
页码：239-250
DOI：10.14257/ijsia.2015.9.11.23
出版社：SERSC
摘要：As network techniques have become one of the most significant aspects of our daily lives, network security has been a major concern. One common application is network intrusion detection. From the perspective of data oriented consideration, intrusion detection can be formulated as a clustering task, which aims to differentiate normal and insecurity behaviors and categorize into several groups. In this paper, we employ ensemble clustering method to improve the generalization and robustness of basic clustering. Specifically, we employ fuzzy kernel C-means (FKCM) as basic clustering, which improves the fuzzy C-means (FCM) clustering by introducing kernels from the support vector machines (SVM) to optimize the features of sample data by mapping the sample pattern into a higher dimensional feature space. Then, we formulate the ensemble problem as the optimization of the mutual information among all clusterings and introduce Ant Colony Optimization (ACO) as the solution. Experiments prove the efficiency of our method.
