Characterizing temporal anomalies in evolving networks

Abstract

Many real world networks evolve over time indicating their dynamic nature to cope up with the changing real life scenarios. Detection of various categories of anomalies, also known as outliers, in graph representation of such network data is essential for discovering different irregular connectivity patterns with potential adverse effects such as intrusions into a computer network. Characterizing the behavior of such anomalies (outliers) during the evolution of the network over time is critical for their mitigation. In this context, a novel method for an effective characterization of network anomalies is proposed here by defining various categories of graph outliers depending on their temporal behavior noticeable across multiple instances of a network during its evolution. The efficacy of the proposed method is demonstrated through an experimental evaluation using various benchmark graph data sets. © 2014 Springer International Publishing.