入侵检测中基于遗传禁忌搜索的模糊聚类的应用

Novel improved fuzzy clustering algorithm applied in network intrusion detection

传统的模糊C均值聚类(FCM)算法须事先指出聚类数,该算法对孤立点和初始聚类敏感、易陷入局部最优,这些因素都将影响最终聚类结果的质量。针对这些缺陷,采用遗传算法和禁忌搜索的混合策略对FCM进行改进,该策略兼具了这两种算法的优势,改进后的算法自动生成最佳聚类数,优化初始聚类的选择,增强算法的爬山能力,有效改善了算法的性能。将改造前后的两种算法用于网络入侵检测实验,实验结果表明,改造后的算法产生的聚类质量明显优于原算法,用新算法对入侵检测建模,提高了模型的自适应性和实用性。

When using traditional fuzzy C-Means（FCM）,the number of clusters must be given beforehand.Furthermore,the algorithm is sensitive to the isolated data and the original clusters and easy to run into local critical point,and these factors have a great influence on the quality of the final clustering results.Because of the faults,the text uses a mixed search strategy which combines genetic algorithm and tabu search to improve the efficiency of FCM.The strategy possesses the advantages of these algorithms.The purpose of this mechanism is to produce the best number of clusters automatically,optimize the selection of the original cluster and advance FCM＇s ability of breaking away from local critical point.Experiments show that the improved algorithm,with self adaptiveness and high efficiency,gets better clustering results than the original.