SHARC: Community-based partitioning for mobile ad hoc networks using neighborhood similarity

Abstract

In this contribution, we present SHARC, a Sharper Heuristic for Assignment of Robust Communities. This algorithm performs distributed network partitioning into communities using epidemic propagation of community labels and the computation of a neighborhood similarity metric. Due to its decentralized nature, SHARC is scalable and well suited for networks where no global knowledge nor node coordination exist, like ad hoc networks. Besides, SHARC is computationally efficient and does not depend on configuration parameters. We validated our approach and compared it to alternative solutions using static and dynamic networks. Results show that SHARC provides a sharper and more robust community assignment and prevents the domination of a single community in both static and dynamic networks.