Title:Distributed Community Detection in Dynamic Graphs

Abstract:The social behaviour of human agents in Opportunistic Networks has a strong impact on the performances of communication protocols since the dynamic topology of such networks is often conditioned by the presence ofCommunities, i.e., sets of agents that use to meet each other much more often than the average intermeeting time. Community Detection is thus a preliminary task that may reveal to be crucial in order to design and analyse efficient communication protocols for opportunistic networks. Our contribution is the first analytical study of this task in dynamic networks. We provide a framework that formalizes the distributed version of the Community-Detection Problem over a general model of dynamic networks. According to this framework, the problem turns out to be a node-coloring task of the dynamic graph. Then, we present an efficient provably-good coloring protocol for two classes of dynamic random graphs that have been recently adopted as mathematical models of some opportunistic networks.