Outlier Detection in Graph Streams Charu C. Aggarwal ∗1, Yuchen Zhao #2, Philip S. Yu #3 ∗IBM T. J. Watson Research Center Hawthorne, NY 10532, USA 1
[email protected] #University of Illinois at Chicago Chicago, IL, USA
[email protected] [email protected] Abstract—A number of applications in social networks, may be considered an edge in the graph stream. The telecommunications, and mobile computing create massive pattern of interactions within a small time window, or streams of graphs. In many such applications, it is useful to detect within users of a particular type may be considered a structural abnormalities which are different from the “typical” behavior of the underlying network. In this paper, we will provide structural network stream. Similarly, certain events in first results on the problem of structural outlier detection in mas- social networks may lead to local patterns of activity, sive network streams. Such problems are inherently challenging, which may be modeled as streams of graph objects. because the problem of outlier detection is specially challenging • The user browsing pattern at a web site is a stream of because of the high volume of the underlying network stream. graph objects. Each object corresponds to the browsing The stream scenario also increases the computational challenges for the approach. We use a structural connectivity model in pattern of a particular user. The edges represent the path order to define outliers in graph streams. In order to handle the taken by the user across the different objects. sparsity problem of massive networks, we dynamically partition the network in order to construct statistically robust models of The afore-mentioned examples typically correspond to sce- the connectivity behavior.