Angling for Big Fish in BitTorrent

BitTorrent piracy is at the core of fierce debates around network neutrality. Most of the legal actions against BitTorrent exchanges are targeted toward torrent indexing sites and trackers. Surprisingly, little is known about the initial seeds that insert contents on BitTorrent and about the highly active peers that are present in a large number of torrents. The main reason is that acquiring this knowledge requires a large-scale continuous crawl of BitTorrent, which is believed to be impractical. However, this information is important for scientist and politics as many unfounded claims are made about BitTorrent piracy. In this paper, we present a crawl dedicated to initial seeds identification and a large-scale continuous crawl of 103 days during which we collect 148M IP addresses of peers participating in 1.2M of torrents. We present the first in-depth analysis of initial seeds' behavior and of highly active peers. We show that it is possible to identify initial seeds for 70\% of torrents, that initial seeds form a small community, and that some of the most active initial seeds are in hosting centers, which make the identification of the location of the human being running those initial seeds complex. In addition, we identified among the highly active peers very different categories including anti-piracy groups and VPNs like iPREDator. We also confirmed that Tor is inefficient for BitTorrent distribution and that VPN solutions are used by well-provisioned peers. Finally, we found an issue affecting initial seeds in the way torrents are queued in BitTorrent clients