Forgot your password?
typodupeerror
Privacy

Digital Watermarks to Replace DRM 374

Posted by CmdrTaco
from the now-we're-getting-somewhere dept.
palegray.net noted a wired story about an industry trend towards watermarking and away from DRM. It says "With all of the Big Four record labels now jettisoning digital rights management, music fans have every reason to rejoice. But consumer advocates are singing a note of caution, as the music industry experiments with digital-watermarking technology as a DRM substitute. Watermarking offers copyright protection by letting a company track music that finds its way to illegal peer-to-peer networks. At its most precise, a watermark could encode a unique serial number that a music company could match to the original purchaser. So far, though, labels say they won't do that: Warner and EMI have not embraced watermarking at all, while Sony's and Universal's DRM-free lineups contain "anonymous" watermarks that won't trace to an individual." Here is a Technical discussion on AudioBox and PSU.edu's Abstract Index
This discussion has been archived. No new comments can be posted.

Digital Watermarks to Replace DRM

Comments Filter:
  • Watermarks (Score:1, Informative)

    by saibot834 (1061528) on Saturday January 12, 2008 @01:43PM (#22015926) Homepage
    Watermarks are still DefectiveByDesign [defectivebydesign.org]
  • by Asmor (775910) on Saturday January 12, 2008 @02:00PM (#22016170) Homepage
    No...

    Let's take a sentence as an example:

    The quick brown fox jumped over the lazy dog.

    Now, let's watermark that sentence for a few different people.

    j498fn894The quick brown fox jumped over the lazy dog.
    j89g5m6-0The quick brown fox jumped over the lazy dog.
    iebciemgtThe quick brown fox jumped over the lazy dog.

    By comparing each of those sentences, you see the first few characters are different in each, thus you can assume that's where the watermark is.
  • by Anonymous Coward on Saturday January 12, 2008 @02:06PM (#22016250)
    Uh, there are much much more clever ways to watermark a file for which such a comparison would be useless. The method you describe is absurdly basic.
  • by spitzak (4019) on Saturday January 12, 2008 @02:16PM (#22016364) Homepage
    In the real watermarking scheme, every single byte is changed. Basically the entire thing is covered with a huge watermark that is noise, with randomly and sparsely distributed blocks of the actual watermark. So finding identical bytes does not work.

    Averaging would seem to work but supposedly the algorithims can survive quite a lot of coverage with random noise. If the watermarks are sparse enough, all that averaging will do is make a result that has *all* the watermarks of the originals. What they do need to do is avoid having huge numbers of different watermarks, as I doubt it will survive tens of thousands of different samples being averaged. This is probably a reason there will not be per-user watermarks.
  • by CrazedWalrus (901897) on Saturday January 12, 2008 @02:19PM (#22016422) Journal
    I would think it's possible to detect using two watermarked as well.

    Think about this as a watermark:

    1234JustShootMe567
    1234CrazedWalrus567

    1234567 doesn't identify anyone, and I've found and removed the portion of the code that differentiates you or me. If the watermark is tied to the user, then that part of it is necessarily different. This assumes that the file is not re-encoded for every user before adding the watermark. Doing so would be a major detriment to scalability, so I doubt that could be done.

    Even if it is encrypted, it would have to be placed in an area of the music that isn't significant -- maybe a least-significant-bit of one channel or something -- or you'd hear it. If that's the case, then if you have two files from two different users, you can bitwise-or, zero-out, or otherwise destroy the information wherever the bits differ between the files. Since they're necessarily in an insignificant part of the signal, the music probably won't sound noticeably different.

    I just think this sounds incredibly weak. If people can break encryption and decode entire streams, there are going to be ways to strip these watermarks -- probably the day the first song that contains it is released.
  • by ceoyoyo (59147) on Saturday January 12, 2008 @02:30PM (#22016532)
    Did you read that? They talk mostly about steganography and watermarks and note that the information can be robust to one transformation but is generally destroyed by several. They mention fingerprinting, which is really what this thread is about, where a different watermark is applied to every copy, but they certainly don't imply that it's not removable or undetectable. The chart at the beginning suggests that it would be desirable if the fingerprint was not removable.
  • by hedwards (940851) on Saturday January 12, 2008 @02:40PM (#22016650)

    Then you're screwing up the music with noise and the people buying it would be better off just to download a copy ripped from CD. You're also loading up your servers with a huge amount of processing required to produce a different version of the entire file for each customer.
    Yes, but when they've been seriously worked on, the result isn't noticeable to anybody not comparing checksums.

    The technique is based upon steganography, and it also works better in higher quality files than in the 126 bitrate junk. Nobody hears everything in a sound file once there is enough complexity, and the watermark parts go into the areas that people aren't able to really hear.

    There's no reason why an end user, or anybody other than the person doing the watermarking needs to be able to find it. If you randomly intersperse the watermark through a large enough portion of the file, it becomes quite difficult to find and effectively remove without causing damage to the file.

    The trick to it is to touch every single frame, but in random spots, and to do so with enough variety that you would need to compare a huge number of copies to have a shot at unwatermarking the file. Doing so will change the results of the checksums making it a pain to figure out where the signature actually belongs. Most of the changes don't even have to have anything to do with the watermark. The weakness then is comparing against a clean copy, and to be honest, anybody that has a clean copy and cares about the watermarking is just going to use the clean copy. And if there's enough variability, it's going to be a tough thing to strip out without causing other problems.

    It's one of those things where unless you've allowed your copy to make it onto the net, nobody is going to be able to examin the file anyways. It is several steps above the current system in terms of convenience. One could probably screw it up by transcoding it, but that is similar to what ITMS allows presently, and it does lose quality as well.
  • by Anonymous Coward on Saturday January 12, 2008 @02:41PM (#22016654)
    That won't work.
    Good audio watermarks can survive 64Kb/s mp3 encoding with an added 50db of white noise.
    They don't care about the huge error rates as the amount of embedded information is very small, in the order of 40 bytes or so, and you have on average three minutes of signal to get one single recovered tiny block of data.

    You can screw with it as much as you like, but it's impossible to remove the watermark without destroying the audio.
    Don't think of it as a succession of samples that can be compared. Think about it, how can you find a tone by looking at single samples? Pseudo random wandering frequency tones are one of the methods used.

    Also, you need to be able to tell if you were successful in removing it, which is impossible without the decryption tools.
  • Not Enough Credit (Score:4, Informative)

    by raftpeople (844215) on Saturday January 12, 2008 @02:44PM (#22016686)
    While I agree that people are going to break this, I don't think you are giving enough credit to the engineers and acedemics working on this problem. Read this linked article and you will see that it's far more complex than just bit twiddling (although clearly there will be differences in the bits ultimately).
    http://209.85.173.104/search?q=cache:kNuSjbUY1iYJ:www.fxpal.com/publications/FXPAL-PR-03-212.pdf+watermarking+audio&hl=en&ct=clnk&cd=1&gl=us [209.85.173.104]
  • Re:Give and Take (Score:5, Informative)

    by ColdWetDog (752185) on Saturday January 12, 2008 @02:45PM (#22016706) Homepage

    Most people are calling for some reasonable give and take, in that regard i cant really argue against watermarks.

    You might want to argue against watermarking technology if you'd had RTFA.

    Digital audio watermarking involves the concealment of data within a discrete audio file. Applications for this technology are numerous. Intellectual property protection is currently the main driving force behind research in this area. To combat online music piracy, a digital watermark could be added to all recording prior to release, signifying not only the author of the work, but the user who has purchased a legitimate copy. Newer operating systems equipped with digital rights management software (DRM) will extract the watermark from audio files prior to playing them on the system. The DRM software will ensure that the user has paid for the song by comparing the watermark to the existing purchased licences on the system.
    (emphasis mine)

    TFA goes on to describe how this is a bit difficult in practice with current technology, but "they're working on it". Given the hit that classic DRM is taking in the PR space now, and given that the media company execs haven't all dropped acid and wandered back into the sixties, I think it's a safe bet that they're going to work on DRM II (New and improved, patent pending). You may return to wearing your tin foil hats now.

  • Transcoding (Score:3, Informative)

    by flyingfsck (986395) on Saturday January 12, 2008 @02:51PM (#22016786)
    I'd love to see an inaudible/invisible water mark that can survive transcoding from MP3 to ogg, or from MPEG4 to MJPEG for example.
  • by Firethorn (177587) on Saturday January 12, 2008 @03:06PM (#22016966) Homepage Journal
    Even after you've averaged three copies together?

    Only what at least two copies have in common would remain intact.

    We're talking bit comparisons here - massive redundancy gains you nothing.
  • by ameoba (173803) on Saturday January 12, 2008 @03:28PM (#22017238)

    I think a better example might be like...

    1. The quick brown fox jumped over the lazy dog
    2. The quick brown fox jumped over the lazy dog
    3. The quick brown fox jumped over the lazy dog
    4. The quick brown fox jumped over the lazy dog
    5. The quick brown fox jumped over the lazy dog
    6. The quick brown fox jumped over the lazy dog



    (view source if you don't see it)
  • by Anonymous Coward on Saturday January 12, 2008 @04:09PM (#22017626)
    more likely you'll retain artifacts of BOTH watermarks and be doubly guilty. furthermore, depending on the watermarking scheme, you could wind up with a significantly degraded file.
  • Re:Transcoding (Score:3, Informative)

    by swilver (617741) on Saturday January 12, 2008 @05:18PM (#22018304)
    Since the purpose of most lossy audio compression algorithms is to make the audio as small as possible by removing as many stuff you can't hear (ie, unheard information) then the inevitable outcome is that at some point watermarks must be audible or the latest codec will strip it.
  • by ASBands (1087159) on Saturday January 12, 2008 @06:29PM (#22019034) Homepage

    Theoretically, the watermark should still be there, as the watermark is inaudible noise on the track. The goal of a good watermarking algorithm is to survive longer than the audio. You're not safe from detection by transcoding, as these guys [igd.fhg.de] have an algorithm (I'm sure many more do, as well) for the original audio track (off a CD) that can be "retrieved" at various bit rates. At the bottom, you can see a graph on the error rate of recovery, which doesn't really fall off until you get down to 64 kbps. Basically, to remove this watermark without knowing the key (which can be as large as needed), you would do more damage to the sound of the track than the background noise.

    As long as a strong watermarking scheme is used, it will still be there, unless you screw up the sound. I don't think it will do anything for the RIAA, but it beats the hell out of DRM and root kits.

  • by Ossifer (703813) on Saturday January 12, 2008 @06:40PM (#22019150)
    If the RIAA were to attempt to utilize watermarking in court, then the watermarking algorithms would become fair game for the defense, which would yield to publicly available methods to strip the watermarking...
  • by Anonymous Coward on Saturday January 12, 2008 @08:12PM (#22019932)
    I'm from Germany, you insensitive clod.

    I pay a fee every time I buy an audio CD, a blank CD-ROM, or even a CD writer; at the same time, I have a legal RIGHT to make copies of my CDs for my friends (which is why the government allows those agencies to collect the fee in the first place).

    No, arguing that everyone I'm sharing files with on the Internet is my friend won't work; people have tried that. But making unauthorised (!) copies of CDs for my actual friends is absolutely and explicitely legal, and, given that I'm paying for it, not immoral in the slightest bit, either.
  • by Anonymous Coward on Sunday January 13, 2008 @04:57PM (#22027896)
    All whitespace gets collapsed into a single space, according to the HTML standard.

APL hackers do it in the quad.

Working...