Decentralized Consensus for P2P Network with Trust Relationships

Bitcoin presents the application of reaching consensus over an open p2p network in a decentralized way. To circumvent the limitations of existing approaches, we proposed a solution based on trust relationships and social consensus. We start from the modeling and analysis of the synchronous consensus algorithm, which is impractical but convenient to get insight of the characteristics of the algorithm, and then adapt it to an asynchronous one, which is practical in a real p2p network. We focus on the convergence and collusion attack of the algorithm. Simulations show that on the SNAP dataset of the Wikipedia who-votes-on-whom network, the synchronous algorithm can always converge within 40 rounds, and the asynchronous algorithm can always converge in about 70 seconds under reasonable latency assumption. Simulations also show that the algorithm can also tolerance collusion attack of over 15% random nodes and 2% top influential nodes.