关键词: 网络入侵检测 无导师学习 核聚类 模式挖掘

Abstract:

Based on the framework of network intrusion detection systems based on data mining , this paper devises an analyzer model of unsupervised learning. It  applies the kernal k-means clustering, pattern mining and approximate pattern matching techniques. It does not use the tag data in the training procedu  re, and need not to ensure the amount ratio of normal data and exceptional data in each training data set. The experimental results show that the model   features high detection rates and low false-positive error rates.

Key words: network intrusion detection, unsupervised learning, kernal clustering, pattern mining