| Awerbuch and Scheideler have shown that peer-to-peer overlays networks can only survive Byzantine attacks if malicious nodes are not able to predict what will be the topology of the network for a given sequence of join and leave operations. A prerequisite for this condition to hold is to guarantee that nodes identifiers randomness is continuously preserved. However targeted join/leave attacks may quickly endanger the relevance of such an assumption. Inducing churn has been shown to be the other fundamental ingredient to preserve randomness. Several strategies based on these principles have been proposed. Most of them are based on locally induced churn. However either they have been proven incorrect or they involve a too high level of complexity to be practically acceptable. The other ones, based on globally induced churn, enforce limited lifetime for each node in the system. However, these solutions keep the system in an unnecessary hyper-activity, and thus need to impose strict restrictions on nodes joining rate which clearly limit their applicability to open systems. In this paper we propose to leverage the power of clustering to design a provably correct and practically usable solution that preserves randomness under a bounded adversary. |
Récupérer au format
