Autonomic detection of dynamic social communities in Opportunistic Networks

In this paper we focus on approaches which aim at discovering communities of people in Opportunistic Networks. We first study the behaviour of three community detection distributed algorithms proposed in literature, in a scenario where people move according to a mobility model which well reproduces the nature of human contacts, namely HCMM. By a simulation analysis, we show that these distributed approaches can satisfactory detect the communities formed by people only when they do not significantly change over time. Otherwise, as they maintain memory of all encountered nodes forever, these algorithms fail to capture dynamic evolutions of the social communities users are part of. To this aim we propose ADSIMPLE, a new solution which captures the dynamic evolution of social communities. We demonstrate that it accurately detects communities and social changes while keeping computation and storage requirements low.