摘要
探索一种基于聚类来识别异常的方法,这个方法不需要手动标示的训练数据集却可以探测到很多不同类型的入侵行为.实验结果表明该方法是可行的和有效的,使用它来进行异常检测可以得到探测率和误报率的一个平衡,从而为异常检测问题提供一个较好的解决办法.
In this paper we explore a clustering algorithm to identify outliers. No manually classified data is necessary for training and it is able to detect many different types of intrusion. The experiment result shows that the algorithm is feasible and effective, and anomaly detection by using this algorithm could get a balance between false positive rate and detection rate, .so it could be a better solution to anomaly detection.
出处
《微电子学与计算机》
CSCD
北大核心
2008年第5期62-65,共4页
Microelectronics & Computer
关键词
入侵检测
聚类
特征检测
异常检测
intrusion detection
clustering
signature detection
anomaly detection
