Abstract: The distance definition for mixed attribute, the outlier factor measured outlier degree of an object and unsupervised intrusion detection method are proposed. The method has the nearly linear time complexity with the size of dataset and the number of attributes. This results in good scalability. The method is robust to the parameter and the input sequence of data, also can detect unknown intrusions.

摘要： 提出了一种适用于混合属性的距离定义和度量对象异常程度的异常因子,由此提出了一种无指导的入侵检测方法。该方法关于数据库的大小、属性个数具有近似线性时间复杂度,这使得算法具有好的扩展性。该方法对于参数和数据输入顺序具有稳健性,且对新的入侵有一定的检测能力。

关键词: 入侵检测, 聚类, 异常因子

CLC Number: 

• TP393.08