On the efficacy of detecting and punishing selfish peers

  • Authors:
  • Byung-Cheol Kim;Jonathan K. Shapiro

  • Affiliations:
  • Dept. of Economcis, Michigan State University, East Lansing, MI;Dept. of Computer Science and Engineering, 3115 Engineering, Michigan State University, East Lansing, MI

  • Venue:
  • WINE'05 Proceedings of the First international conference on Internet and Network Economics
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

We study the phenomenon of free-riding in peer-to-peer (P2P) systems via an abstract model of a public good provision game with incomplete information. Each user in the model has an intrinsic contribution cost and decides whether to contribute or free-ride based on the expected benefit derived from the system. We first consider the impact of positive network externalities—common in P2P settings—on the equilibrium level of free riding and show that such network effects reduce free riding slightly but are insufficient to prevent it. We then consider the use of an incentive mechanism based on the detection and punishment of selfish peers, explicitly modelling key design tradeoffs inherent in any realistic detection mechanism. We show that detection and punishment can reduce free riding, but that the risk of falsely accusing cooperative peers can diminish their effectiveness. Finally, we consider what level of detection would maximize the social welfare of the network. We find that a surprisingly low level of detection can be optimal and that the residual level of free riding at optimum depends critically on the overhead of detecting selfishness and the probability of falsely identifying cooperative peers.