Impact of Node Cheating on Gossip-Based Protocol

摘要

Gossip-based protocol has been widely adopted by many large-scale multicast applications. In this paper, we study the impact of node cheating on decentralized gossip-based protocol. We mainly focus on two cheating strategies, one is to increase the subscription request sending times, and the other is to increase the PartialView size. We establish a cheating model of nodes for gossip-based protocol, and evaluate the system performance when node cheating happens. In the simulations, we analyze the impact of node cheating on a representative gossip-based protocol, SCAMP (Scalable Membership Protocol, a decentralized, gossip-based protocol). The results show that node cheating makes considerably negative effects on the system performance, and there exists a delicate relationship between the percentage of cheating nodes in the system and the benefit they can gain. The study results also show that cheating behaviors should be paid much more attention during the gossip-based protocol design in future.