GNAED: A data mining framework for network-wide abnormal event detection in backbone networks

In backbone networks, due to the frequent exchange of data between adjacent routers, the characteristics of network behaviors caused by network distributed abnormal events are closely related with routers´ spatial location, their connection relationships and other associated information. Based on this observation, this paper presents a novel graph-based abnormal event detection framework that uses a set of data mining techniques for facilitating the detection of distributed abnormal events in backbone networks. The experiment results using real traffic data collected from an ISP backbone network show the effective and scalable of the proposed framework.