EigenPulse: Detecting Surges in Large Streaming Graphs with Row Augmentation.

How can we spot dense blocks in a large streaming graph efficiently? Anomalies such as fraudulent attacks, spamming, and DDoS attacks, can create dense blocks in a short time window, emerging a surge of density in a streaming graph. However, most existing methods detect dense blocks in a static graph or a snapshot of dynamic graphs, which need to inefficiently rerun the algorithms for a streaming graph. Moreover, some works on streaming graphs are either consuming much time on updating algorithm for every incoming edge, or spotting the whole snapshot of a graph instead of the attacking sub-block.