On the detection of cheating on private BitTorrent trackers

Abstract

BitTorrent has become the most important protocol for peer-to-peer file sharing, and BitTorrent-related communication makes up a significant fraction of the total traffic on the Internet. The incentive mechanism to ensure a fair exchange of data between peers is a ``tit-for-tat'' variant, a mechanism that rewards other users connected to the same swarm for uploading, by reciprocally uploading to them. Unfortunately, as previous work has shown, this mechanism is not only largely ineffective against selfish peers, but it also does not help against ``hit-and-run'' peers (peers that disconnect once they have completed their download), which leads to the starvation of torrents. Private BitTorrent trackers attempt to solve the problems of misbehaving peers by requiring their users to maintain a certain ratio between uploaded and downloaded traffic. However, the tracker has to rely on traffic statistics that the peers provide themselves. As a result, malicious peers can misrepresent the traffic volume that they have exchanged, refuse to report back, or pretend not to be connected at all. This results in problems similar to the ones faced by public trackers.<br /> In this thesis, we first present the inner workings of private BitTorrent trackers and the way in which they keep track of user traffic statistics. Then, we discuss attack techniques used by malicious peers (cheaters) to download more data than they are entitled to and dissect them in turn. We then focus on cheaters, who spoof uploads and we present two techniques to identify cheaters in swarms.<br /> The first technique checks for the consistency of the traffic volumes that peers report within a single BitTorrent swarm. The second technique aggregates and correlates the activity of peers when they are active in multiple swarms. Our evaluation shows the effectiveness of both approaches, even in the face of countermeasures cheaters might employ to stay undetected, before concluding with a discussion of related work and an outlook on the future of cheating in BitTorrent.