Change Detection in Large Evolving Networks

Abstract

This article presents a novel technique for the detection of change in massive evolving communication networks. This approach utilizes a novel hybrid sampling methodology to select central nodes and key subgraphs from networks over time. The objective is to select and utilize a much smaller targeted sample of the network, represented as a graph, without loss of any knowledge derived from graph properties as compared to the entire massive graph. This article uses the targeted samples to detect micro- and macro-level changes in the network. This approach can be potentially useful in the domain of cybersecurity where this article highlights the importance of graph sampling and multi-level change detection in identifying network changes that may be difficult to detect on a larger scale. This article therefore presents a means to audit large networks to establish continuous awareness of network behavior.