Targeted and scalable information dissemination in a distributed reputation mechanism

In online reputation mechanisms, providing the system participants (peers) with the appropriate information on previous interactions is crucial for accurate reputation evaluations. A naive way of doing so is to provide all peers with all information, regardless of whether they need it or not, which may be very costly and not scalable. In this paper we propose a similarity-based approach, named SimilDis, for targeted dissemination of information in the distributed reputation mechanism called BarterCast. In BarterCast, each peer collects information on the interactions (data transfers) that have occurred in the system, and builds a weighted directed graph that represents its partial view of the system. We propose two methods to derive peer similarity in the partial graph of a peer. The first method is based on incrementally maintaining a directed acyclic graph, and the second method is based on performing multiple nonuniform random walks in the partial graph. In both methods, each peer maintains a list of the peers most similar to itself, and gives higher priority to them when disseminating information. We evaluate the accuracy and the cost of these methods using trace-driven simulations based on traces from the Tribler P2P file-sharing network, which employs BarterCast. As the results show, both methods exhibit very small errors in the computed reputations in comparison with the case of providing complete knowledge to all peers, but decrease the communication and storage costs by two orders of magnitude.