摘要
在网络流量管理中流量异常的一般检测方法是阈值监控。文章提出一种新的异常检测方法,选取适当的SNMP管理信息库变量,建立对相关变量的局部AR(自回归)模型,检测并分析一种服务器故障引起的流量异常,获得该故障的特征向量模型;该检测方法比阈值方法有更强的检测功能,并与传统GLR测试方法进行对比。
In general,the traffic anomaly is detected using a threshold in network traffic management.In this paper,a new method was presented to dete ct traffic anomaly.By choosing the relevant variables of the SNMP management information base,modeling their observations with AR(Autoregressive)model, de-tecting and analyzing the traffic anomaly caused by a server fault,and one feature vector model of the fault was composed.This kind of method can detect some detailed anomalies those cannot be detected by threshold,and the compariso n between it and GLR test was given.
出处
《微电子学与计算机》
CSCD
北大核心
2002年第12期1-6,共6页
Microelectronics & Computer
基金
国家重大基金项目(90104006)
