Sleuthing from public sources to figure out how the Hateful Eight leaker was caught
In 2014, Quentin Tarantino sued Gawker for publishing a link to a leaked pre-release screener of his movie "The Hateful Eight." The ensuing court-case revealed that the screeners Tarantino's company had released had some forensic "traitor tracing" features to enable them to track down the identities of people who leaked copies.
Working from court records, as well as documents from the North Korea/Sony hack, a patent lawsuit against Tarantino's forensics company, and the forensic company's own patent filings, Matthew Fuller and Nikita Mazurov do an incredible job of sleuthing to uncover the interior, technical workings of the secretive world of digital copyright-enforcement forensics.
The paper itself is published in the journal Theory, Culture and Society and frames the investigation in the context of cultural studies. It begins with a lot of theory that I found to be quite a struggle, but the technical detail (which is on pages 9-18) is really fascinating.
Fuller and Mazurov use their technical analysis to make some notes towards a "counter-forensics" of techniques that would defeat the traitor-tracing measures.
This reminds me of the work Ed Felten and co did on SDMI more than a decade ago, in which they presented the general hypothesis that watermarks could be robust (hard to remove) or imperceptible (not so obtrusive as to wreck the user's enjoyment) but not both. If a watermark adds no perceptible data to the signal, then it can be removed with no perceptible loss -- and since you can compare two or more copies of a work to find the watermarks, it's never that hard to find the marks in order to remove them.
Recalling that Deluxe's 'SecureCinema Digital Screener Platform' presentation mentions that its FCT watermarking technology is patented, a search was performed to identify possible patents and patent applications filed by Deluxe or its subsidiaries. Two relevant patents were found during this discovery stage. The first patent application submitted, filed in 2003 and published in 2005, appears to discuss the aforementioned standard mode of coded symbol-based visual water-marking, a method for incorporating into film frames 'images or patterns that appear as unobtrusive defects or artifacts', specifically constituting 'a pattern of small, unobtrusive specks'. However, as the Deluxe presentation explicitly stated that 'no visual artifacts are added to picture', it would seem that this was an early prototyping of Deluxe's watermarking technique, as opposed to a mechanism currently in use, at least under the FCT banner. A patent filed in 2004 and published in 2006 and reissued in 2015, proposes a video watermarking scheme in which objects in a given frame are themselves augmented, wherein '[i]tis preferred to do this by enlarging an image slightly so that one or more edges of the image is moved relative to the same edge in the video master'. Given that this technique is in accord with the dictum that no visual artifacts are added to the video, as only existent images are manipulated, it would thus seem that this patent is that covering the 'FCT Film' components of Deluxe's FCT watermarking schema.Recalling the mention of the server name in the sample Web Watch report, the patent further notes that the altered (watermarked) video is stored on a given 'modification' server after it is encoded and water-marked from the master copy on the master server. Thus, the aforementioned server name in the Web Watch report may presumably identify which modification server the given copy of the film was stored on.A second patent, filed in 2007 and issued in 2008, deals with the 'FCT Sound' component of Deluxe's watermarking scheme and proposes a method for watermarking audio tracks notvia the addition of extraneous audio artifacts, but via the removal of existent ones. Specifically, 'the analog soundtrack is altered by selectively muting portions of the analog soundtrack at the selected location for the insertion of the identifiable code'. Whilst the patent literature is thus more explanatory than Deluxe's official web-facing material and news sources, more explication still maybe found in legal proceedings. In 2010, the Swiss company Medien Patent Verwaltung AG filed a complaint in US courts stating that Deluxe had infringed on its (American) anti-piracy watermarking patent, alleging that Deluxe had manufactured film prints which employed Medien's anti-piracy techniques. Amongst other documents Deluxe filed a reply memorandum contesting Medien's claims which was in turn denied by the court , centering around the fact that since its audio water-marking system functioned around the subtraction of information not its addition, that its own patent did not infringe upon Medien's patent which only discussed 'markings', not their removal. During the course of the various court dockets, however, Deluxe divulged further information about the inner workings of its FCT sound watermarking procedure, including sample images of water-marked film prints with portions of the film's audio track being obfuscated to create 'mutes or micro-second cancellations' . In another court docket Deluxe crucially revealed that '[i]mportantly, these codes are hidden in sound effects so that they are not noticeable to the audience'). The information extracted from these legal proceedings reveals that the FCT sound watermarks operate via the deletion of micro-second durations of parts of a film's audiotrack, and may likely occur during sound effects in the soundtrack, so as to attempt to mask their perceptibility.
A Counter-Forensic Audit Trail:Disassembling the Case of The Hateful Eight [Matthew Fuller and Nikita Mazurov/Theory, Culture and Society]
(Thanks, Matthew!)