一种云环境下的高效异常检测策略研究

AN EFFICIENT ANOMALY DETECTION SCHEME IN CLOUD ENVIRONMENT

针对虚拟机进行异常检测是提高云计算系统可靠性的重要手段之一。然而,云环境中虚拟机的性能指标数据具有维度高、信息冗余等特点,会降低检测效率和准确度。同时,传统异常检测方法难以定量刻画系统的异常状态,而局部异常因子(Local Outlier Factor,LOF)算法虽可量化其异常程度,但它以相同权重计算不同维度变量对系统状态的影响,导致算法对异常的区分能力减弱。针对以上问题,提出一种高效的异常检测策略。该策略以最大相关最小冗余算法和主成分分析法对性能指标进行筛选降维,提高了异常检测的效率;为LOF算法中不同维度的变量赋予不同权重,强化了不同指标对异常的区分度。实验表明,该策略相对于传统异常检测方法,效率和检测率都有显著提高。

Anomaly detection for virtual machine is one of the most important methods to improve the reliability of the cloud computing systems.However,performance metric data of virtual machine in cloud environment has the characteristics of high dimension and information redundancy,which will reduce the detection efficiency and accuracy.Meanwhile,it is difficult for the traditional anomaly detection algorithm to quantify the abnormal state of the system.The Local outlier factor(LOF)algorithm can be used to quantify the abnormal degree,but it uses the same weight to calculate the influence of different dimension variables on the system operation state,which leads to the weakening ability of the algorithm to distinguish the anomaly.To cope with the above problems,an efficient anomaly detection strategy is proposed in this paper.The strategy adopted the maximum correlation and minimum redundancy algorithm and principal component analysis method to screen and reduce the dimension of performance metrics,which improved the efficiency of the anomaly detection.It assigned different weights to the variables of different dimensions in the LOF algorithm,and strengthened the discrimination between different indicators.Experimental results show that compared with traditional anomaly detection methods,the efficiency and accuracy of our strategy are significantly improved.