Sharing Incentives for Lazy Free-Riders in BitTorrent and the BarterCast Reputation System

A well-known problem in peer-to-peer networks is free-riding, where users do not share resources in return for what they consume. Free-riders can be distinguished in two categories: diehard free-riders that are willing to subvert the network’s protocol in order to free-ride, and lazy free-riders that are reluctant to share but do follow the protocol. An important body of research focuses on die-hard free-riders in the popular BitTorrent file-sharing network, but in practice diehard free-riding in BitTorrent is not often observed. Lazy free-riding, on the other hand, is often observed, and in this thesis we investigate whether BitTorrent provides lazy free-riders with an incentive to share. Based on a game-theoretical model, we prove that this is the case for some lazy free-riders, but not for all. We then proceed to investigate the same for BarterCast, a new distributed reputation mechanism that is added to the BitTorrent-based Tribler network to provide additional sharing incentives. Based on an extended version of the same model, we prove that BarterCast also provides incentives only to some lazy free-riders, but not for all. We verify these results with simulations, and find that in practice, even fewer incentives are given than our model predicts. However, we show that lazy free-riding can provide a gain but also a loss, and that the net result is difficult to predict, which can be seen as an additional incentive against free-riding.Algorithms GroupComputer ScienceElectrical Engineering, Mathematics and Computer Scienc