Abstract: Abnormal traffic detection is important to network management.This paper aims at the characteristic of multi-hop network，and defines abnormal as traffic-robust after failure.Then the paper presents a fast auto-adaptive traffic-aware robust detection method.This method compresses data via efficient histogram，and detects double-size robust via congregate function.The paper analyzes status of links and node via being aware traffic and can get status of network fast and nicely.It does experiment to validate the validity and advance in performance of this method compared with other robust detection method.

Key words: robust detection, auto-adaptive, traffic-aware

摘要： 网络流量突变检测在网络管理中意义重大，针对多下一跳网络的特点，将异常定义为故障后流量的突变，提出了快速自适应流量感知突变检测算法。算法通过改进的直方图技术压缩数据后，利用聚集函数进行上升和下降两种突变的检测，进而根据感知流量变化来分析网络中节点及链路状态，快速准确地掌握网络整体状况。通过仿真平台对算法有效性进行验证，并与现有突变检测算法进行比较，结果表明该算法能使检测灵敏度和精度得到明显提高。

关键词: 突变检测, 自适应, 流量感知