CLAN: An Efficient Distributed Temporal Community Detection Protocol for MANETs

Real world MANETs often exhibit an inherent community structure in their topological connectivity and in the evolution of the topology over time. Such temporal community structure of MANETs has been shown to be extremely useful in improving the performance of routing and content-based routing in MANETs. However, detecting temporal communities in a completely distributed and real time manner is a hard problem, and it is often performed offline with knowledge of the full network topology over time. We propose CLAN, a distributed and real-time protocol for detecting temporal communities in MANETs. CLAN is an adaptation of the Label Propagation algorithm to distributed and time-varying graphs that MANETs are. A key novel component of CLAN is local rules for community rediscovery as the network evolves. CLAN also uses a weighted version of the network topology where the weights are defined using a novel notion of social entropy to promote stability of communities. Extensive simulation results demonstrate that CLAN is quick to converge, incurs minimal overhead and is as effective as centralized approaches to temporal community detection. We also demonstrate how the temporal community structure can be used by designing a hierarchical routing protocol that achieves the delivery ratio of the OLSR routing protocol at a fraction of the overhead.