基于熵和线性关系的两级流量异常检测方法

Two-Stage Traffic Anomaly Detection Method Based on Entropy and Linear Relation

提出了一种基于熵和线性关系的两级流量异常检测方法,综合考虑了流量异常检测方法的准确性和实时性要求.该方法在时间域上设定两级动态阈值,采用基于熵的方法对异常时间点进行检测,对熵值变化程度明显的时间点可使用一级阈值检测出来,而对熵值变化程度处于一级阈值和二级阈值之间的时间点采用基于线性关系的方法再次进行检测,并通过定义的报警触发函数识别异常类型.仿真实验结果证明,该方法在准确性和实时性方面优于现有的方法.

To improve the accuracy and real-time traffic anomaly detection methods,a two-stage traffic anomaly detection method based on entropy and linear relation was presented. A two-level dynamic threshold was set and used based on entropy and linear relation to detect abnormal traffic in time domain.The type of traffic anomaly can be indentified through the alarm trigger function. Simulations verify that this method is superior to the existing methods in the accuracy and real-time respects.