QPRC 2016

A Random Graph Model for Benchmarking Network Surveillance Techniques

Nathaniel T. Stevens
University of San Francisco

Dynamic networks are often used to model the interactions, or relational structure, of a
group of individuals through time. In many applications, it is of interest to identify
anomalous or spurious behavior among these individuals. The real-time monitoring of
networks for anomalous changes is known as network surveillance. Despite the
prominent development of network surveillance procedures, little to no work has been
done to formally compare existing methods. Because such methods differ widely in their
underlying performance criteria and assumptions, it is important to evaluate and compare
theirs strengths and weaknesses. In this talk we focus on providing and analyzing a
network model that can be used to assess the performance of any network surveillance
method. We use the performance of Shewhart control charts for individuals, under a
variety of simulated conditions, as a benchmark against which other surveillance methods
can be compared.

Keywords: community detection; dynamic graphs; statistical process control; anomaly
detection; online surveillance

Co-collaborator: James D. Wilson, University of San Francisco