Published: June 05, 2008 – 07:50PM CT
A network printer should not be the target of a DMCA takedown notice from US copyright holders, but researchers at the University of Washington have shown just how simple it is to “frame” any particular IP address as an infringing BitTorrent user. The researchers used their technique to attract nearly 500 DMCA takedown notices, all of them bogus and some of them targeting nonsensical devices. Their work shows how difficult it can be to pin down Internet “pirates.”
In a recent paper (PDF), Michael Piatek, Tadayoshi Kohno, and Arvind Krishnamurthy set out to attract bad DMCA takedown notices. In the course of doing earlier research on BitTorrent in 2007, the researchers attracted 206 complaints by accident, even though none of their machines were transferring illegal files. After that experiment ended, they realized that it might be an interesting experiment in itself to try to attract the notices; if they could be hit with 200 bogus complaints, how easy would it be to “frame” someone else on the Internet? Another month of testing brought in 281 more complaints.
Courtesy University of Washington
To understand what the researchers were up to, it helps to understand how content owners police their works on BitTorrent. Trade groups like the RIAA and the MPAA generally farm this work out to specialized companies that monitor public BitTorrent trackers for infringing content. The trackers return a list of peers on the Internet that have at least parts of the file in question.
To bust file-swappers, the companies can take two approaches, “direct” and “indirect” detection. Direct detection involves making an actual connection to the given IP address and downloading part of the file. It’s accurate but resource intensive. The second, and much easier, way to find file-swappers is simply to rely on the IP addresses issued by the tracker. Once these IP addresses are in hand, the companies turn them over to the trade group, which issues either a DMCA takedown notice or (in the case of the RIAA) a “pre-litigation letter” that alerts someone of a pending legal case.
The researchers were able to show that indirect detection is still widely used. By using 13 different machines on the University of Washington campus, the researchers were able to “advertise our presence as a potential replica without uploading or downloading any file data whatsoever.” Despite having no infringing content on their machines, the takedown notices poured in, meaning that “direct detection” could not have been used in these cases.
The huge majority of the complaints involved movies and TV shows; only five takedown notices out of 487 complaints involved music files.
As the experiment showed, BitTorrent tracker data is not always accurate, and in fact can be deliberately planted. Most trackers simply record the IP address of the machine making requests, but other trackers support an extension that allows BitTorrent clients to “specify a different IP address that the trackers should record in its list of peers instead.” While designed to avoid problems with things like proxy servers, this is obviously open to abuse.
The researchers used some of these trackers to “frame” one wireless access point, three networked (and IP-accessible) printers, and a desktop PC that was not currently using BitTorrent at all. Each of the machines received DMCA takedown requests. An attempt to “frame” IP addresses with no machine attached failed to attract any complaints, however.
The researchers conclude that, because of the problems in using “indirect detection,” enforcement agencies “will soon shift to more conclusive methods of identifying users.” But as they do so, resource requirements intensify. Instead of making a simple HTTP request to a tracker, direct detection requires a TCP connection and a block transfer that could range into the hundreds of kilobytes. The researchers estimate that this could mean a 10-100x increase in needed bandwidth.
Their work, they say, isn’t intended to “take sides” in the matter, but simply to show how more transparency is needed about the process of issuing DMCA takedowns. “Without more transparency,” they say, “our concern is that—even if the copyright enforcers fix the problems we’ve identified—there might still be many other flaws remaining and new mistakes introduced in the future.”