关键词: 异常检测, 地球站行为, 聚类分析, 核函数的模糊C均值算法, 最优划分

Abstract: As more and more widely used of satellite communications network，more and more attention is paid on its security.Aiming at possible problems of satellite communications network that may seriously threaten its security，such as lost or falsification of portable earth-station，the idea of anomaly detection of station behavior is brought forword，and by means of combining cluster and pattern recognition technology，an anomaly detection system of satellite communication network is designed and realized.An anomaly detection algorithm is presented based on improved KFCM cluster algorithm which can achieve partial best partition.Experimental results show that the improved algorithm reaches a better clustering result，and the system gets good ability of anomaly detecting.

Key words: anomaly detection, station behavior, clustering analysis, Kenerl Fuzzy C Means（KFCM）, best partition