Abstract

Much work has been done to address the need for incentive models in real deployed peertopeer networks. In this paper, we discuss problems found with the incentive model in a large, deployed peertopeer network, Maze. We evaluate several alternatives, and propose an incentive system that generates preferences for wellbehaved nodes while correctly punishing colluders. We discuss our proposal as a hybrid between TitforTat and EigenTrust, and show its effectiveness through simulation of real traces of theMaze system.

‚Äč