Volume Traffic Anomaly Detection Using Hierarchical Clustering

In a large backbone network, it is important to detect shape traffic fluctuation for servicing robust network. However, there are too many interfaces to monitor the characteristics of traffic. First we collect volume traffic of boundary link. From the volume traffic, we make groups which have similar traffic patterns by hierarchical clustering algorithm. This result shows that most of traffic has similar patterns, but some traffic which is far from centroid has an anomaly traffic pattern. This paper gives a hint for network operators that which traffic has to be checked Out.