中文摘要：

入侵检测数据流具有偏斜分布以及概念漂移的特点,其样本无法准确反映整个空间的数据分布,分类器容易被大类淹没而忽略小类,使得检测正确率不高,对此,文中提出了一种单分类器集成的入侵检测方法,该方法在使用k-means聚类算法调整数据分布的基础上,用区间估计结合AUC的值检查概念漂移并更新分类器。实验结果表明,在处理偏斜数据流上优于均值、乘法规则、最大值三种分类处理方法,并具有较高的入侵检测率。

英文摘要：

We aim to present an algorithm which can obtain higher classification accuracy and higher intrusion detection ratio than those obtainable with the existing algorithms mentioned in the full paper.Section 1 presents our skewed-data stream intrusion detection algorithm.Subsection 1.2 uses the k-means algorithm to adjust data distribution.Subsection 1.3 presents a seven-step procedure for our skewed-data stream intrusion detection algorithm.Our skewed-data stream intrusion detection algorithm uses the interval ...