Robust Incentives via Multi-level Tit-for-tat.

Much work has been done to address the need for incentive models in real deployed peer-to-peer networks. In this paper, we discuss problems found with the incentive model in a large, deployed peer-to-peer network, Maze. We evaluate several alternatives, and propose an incentive system that generates preferences for well-behaved nodes while correctly punishing colluders. We discuss our proposal as a hybrid between Tit-for-Tat and EigenTrust, and show its effectiveness through simulation of real traces of the Maze system. Copyright © 2007 John Wiley & Sons, Ltd.