测度属性关系分析的分布式系统异常检测

Anomaly Detection for Distributed System Based on Measurement Attributes Analysis

针对分布式系统的特点,提出了一种基于网格空间的测度属性关联度计算方法,并以此建立了网络模型来刻画分布式系统测度属性的关系;提出了基于隐马尔可夫模型的测度属性关系网络划分方法;利用最新监测的属性关联度信息,设计了基于测度属性关系网络划分的分布式系统异常检测算法;对多种攻击进行异常检测实验,从而验证了提出方法的有效性.

The anomaly detection methods based on measurement attributes relationship analysis is pro- posed. The correlation between the measurement attributes in space grid structure is calculated, and hereby the relationship network model of measurement attributes is established. The network segmentation method is put forward based on hidden Markov model. The anomaly detection algorithm for distributed systems is designed by using the newest monitoring attributes correlation. The anomaly detection experi- ments demonstrate the effectiveness of the proposed method.