GROUP: A Gossip Based Building Community Protocol

The detection of communities of peers characterized by similar interests is currently a challenging research area. To ease the diffusion of relevant data to interested peers, similarity based overlays define links between similar peers by exploiting a similarity function. However, existing solutions neither give a clear definition of peer communities nor define a clear strategy to partition the peers into communities. As a consequence, the spread of the information cannot be confined within a well defined region of an overlay. This paper proposes a distributed protocol for the detection of communities in a P2P network. Our approach is based on the definition of a distributed voting algorithm where each peer chooses the more similar peers among those in a limited neighbourhood range. The identifier of the most representative peer is exploited to identify a community. The paper shows the effectiveness of our approach by presenting a set of experimental results.