Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: After DRM, watermarking (Read 21533 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

After DRM, watermarking

After several years of failure with DRM, the music industry seems to realise that DRM is not the solution anymore. It's not even that DRM was easily defeated. It was mainly because the consumer didn't want to pay for a format that doesn't really offer the choice of the player/device. So they know now that they should use a more compatible format like mp3, flac, wav, … The problem is that it's not really possible to control the distribution of the content with these format. The new solution would be watermarking. They would use a different watermarking on each tune sold. While this solution doesn't really bother the good consumer, it might be a mean to identify the source of the piracy. It's not even necessary to sue the presumed pirate; you can refuse to sell another track.

There are still some problems with this solution:

First, what's happening if a pirate steals (physically or by hacking) the track you legally bought? I don't think it's a real problem because it won't be difficult to make the difference between the real pirate (who bought the track and systematically redistribute it) and the consumer who was stolen (you can be robbed one or two times, not on each release).

Second, even if it would prevent the piracy for the tune distributed online, it doesn't prevent somebody to buy a CD, rip it and put it on a p2p… It's true but on the long term; the music industry can progressively decide to release their tune only thru online distribution (like twenty years ago the distribution of vinyl was progressively replaced by the CD).

The last "problem" might be the robustness of the watermarking itself. Currently, the only real challenge of the watermarking that I'm aware of is the SDMI contest. Mainly, all the watermarks were defeated by the Professor Felten from the Princeton University. You can read the technical paper from the Pr Felten here: http://www.usenix.org/events/sec01/craver.pdf
However, it was in 2001 and since then a lot of money/work have been invested in the audio watermarking (adaptive watermark, use of echo, …). Nobody was tempted by a new challenge on these new watermarks though…

So here is my question: If the watermarking technology would be widely used, do you think it will be defeated?
I'm not interested by whether it's bad or good. I just want to discuss about the robustness of the solution.

Cheeers,

After DRM, watermarking

Reply #1 –
This has been discussed before. IIRC the points of the discussion were:

- Signatures may survive some lossy compression
- We can design signatures that withstand a specific compression technology
- But a robust signature that survives many processing and withstands all compression technologies may interfere with the quality of the audio

CMIIW...

After DRM, watermarking

Reply #2 –
Like with DRM, I guess that the big guys will either find ways to corrupt watermarks, or they just wont care. Theyll leave the markings assuredly that they cant be traced anyways.

BTW, I read some papers on watermarking in media, and it seems that some hefty signal processing is used for surviving a noisy channel. Maybe Ill get som use for those channel coding lectures after all ;-)

-k

After DRM, watermarking

Reply #3 –
Unfortunately there will always be pirates.  Watermarks, DRM or any other type of added thing built in would be defeated and hacked eventually.  I wish the music industry would concentrate more on providing quality.  Rather than spend millions on R&D for another scheme to stop piracy.  I guess they will never figure it out.

After DRM, watermarking

Reply #4 –
I have not seen ANY software-protection scheme yet, which wasn t hacked, given enough interest and resources. Besides, it would once again just try to fix the symptoms, not the causes.

Financing projects by ENFORCING payment per copy, is dead! The only reason why it is still used, is because of the goodwill of consumers and the ignorance of producers. Copyright is finished - its just that the world is still busy realizing it.

Thus, the correct approach would be to change to different investment models. Interestingly, most of the popular alternatives can be combined:

- funding for the masses: Consumers choose which projects to fund, based on trust. Afterward finishing the original, copies are free.
- pay per original: only applicable to some types of products
- Keep the source, sell generated variations: In some cases, it is possible to create some kind of generator, then keep the generator closed and private and make money by selling INDIVIDUAL products, created by the generator. What you are doing here, basically is selling variations. You own the generator, and sell the stuff which the generator produces.
- sell availability and services: dont make any money by shipping the software at all. Instead, make the software dependend on your service, for which people have to pay. Massive Multiplayer Online Games for example follow this model.
- get rewarded by donations: in the right setting, dogma and a personal environment, people are more willing to reward free of choice without any force necessary. This however is a viable bonus, not something to base your investment on.
- back to the roots: in some cases, good old "i give you this, you give me that"-trading works pretty good. This model however only works if the rules of Mutualism are enforced. Thus, parasites need to get marginalized ruthlessly.

Those are just a few examples. With enough imagination, creativity and flexibility, the death of copyright really isnt such a big issue. Its only a big issue for inflexible and uncreative suckers. Unfortunatelly, the top positions in nowadays industry are mostly filled with this kind of losers.

- Lyx
I am arrogant and I can afford it because I deliver.

After DRM, watermarking

Reply #5 –
My question is not whether it's a good solution or it isn't.
My question is: If the watermarking technology would be widely used, do you think it will be defeated?
Actually, it's more a technical question (that's why I putted it in the the " Scientific/R&D Discussion").
Cheers

After DRM, watermarking

Reply #6 –
I basically answered that already. The problem is that watermarking depends on quality. As we all know, in audio-technology it is easy to destroy something, but nearly impossible to recreate it. In a worst case, simply run some DSP-effects on it, thus distorting and degrading the signal. Given a target to destroy - the watermarking - you can even optimize this process to harm the watermark as much as possible, while being as inaudible as possible to the human ear. How do you imagine to counteract such attacks? In principle, the only way would be to make the watermark audible, so that destroying it requires audible changes to the signal. But who wants to buy music with an audible watermark?

- Lyx
I am arrogant and I can afford it because I deliver.

After DRM, watermarking

Reply #7 –
I red post on HA about this subject and 2bedecided written it wasn't so easy:
http://www.hydrogenaudio.org/forums/index....st&p=465542
However, it was in the context of psychoacoustic compression being able to remove the watermark.

After DRM, watermarking

Reply #8 –
Quote from: pepoluan on
This has been discussed before. IIRC the points of the discussion were:

- Signatures may survive some lossy compression
- We can design signatures that withstand a specific compression technology
- But a robust signature that survives many processing and withstands all compression technologies may interfere with the quality of the audio

CMIIW...



There appear to be at least one or two non-blind watermarks that will survive anything you want to listen to.

S. Kuo, J. D. Johnston, W. Turin, S. R. Quackenbush, “Covert Audio Watermarking Using Perceptually Tuned Signal Independent Multiband Phase Modulation”, ICASSP 2002, Orlando, FL, May 2002

is one, there are others, including at least one followup on that by another author.
-----
J. D. (jj) Johnston

After DRM, watermarking

Reply #9 –
Quote from: Lyx on
Given a target to destroy - the watermarking - you can even optimize
this process to harm the watermark as much as possible, while being
as inaudible as possible to the human ear.

Right, and how do you know that you are harming the watermark
more with transformation B than with transformation A ?

After DRM, watermarking

Reply #10 –
I would argue that generally, the watermark would have to occupy the "signal space" occupied by "percievable audio" in order to survive any conceivable (and non-conceivable) lossy compression based on removing "non-perceivable information".

In practice, there may be enough similarities between current lossy comrpession schemes that a small-bandwidth information snip may be lured in with robustness against compression removal as well as robustness against human detection.

-k

On the other hand, in order to remove or disrupt a watermark sequence with any guarantee, you would either have to change the audio in a perceivable way (degrade it), or be able to identify (and perhaps decode) the sequence and disrupt it non-perceivably.

-k

After DRM, watermarking

Reply #11 –
If you buy several time the same track, can you reject the entropy?

After DRM, watermarking

Reply #12 –
Quote from: knutinh on
I would argue that generally, the watermark would have to occupy the "signal space" occupied by "percievable audio" in order to survive any conceivable (and non-conceivable) lossy compression based on removing "non-perceivable information".

In practice, there may be enough similarities between current lossy comrpession schemes that a small-bandwidth information snip may be lured in with robustness against compression removal as well as robustness against human detection.

-k

On the other hand, in order to remove or disrupt a watermark sequence with any guarantee, you would either have to change the audio in a perceivable way (degrade it), or be able to identify (and perhaps decode) the sequence and disrupt it non-perceivably.

-k


Well, actually, there are redundancies (long, long ones) that audio codecs do not take advantage of. I mean redundancies like 16k sample length redundancies, for instance.

If you use this kind of signalling as pre-processing, you won't ever get a blind detection, but you can get a non-blind detection against the original at remarkably bad quality, perceptual or LMS.

It is concievable you could disrupt these, but if the long-period disruptions are just at the hairy edge of audibility to start with, it's going to be hard to disrupt without making some changes to how the music sounds. (You won't have that original to compare against, and if you did, of course, you'd just use that.)

This, like all methods, is somewhat subject to collusion. If you have 32 different versions, chances are you can sum them, normalize, and disappear the information.
-----
J. D. (jj) Johnston

After DRM, watermarking

Reply #13 –
Quote from: marcan on
If you buy several time the same track, can you reject the entropy?

Most likely so. Unless they changed repurchases in other ways, you'd just have to eliminate the parts which are not in all versions - though distortion may occur
err... i'm not using windows any more ;)

After DRM, watermarking

Reply #14 –
I have to agree with Lyx:

The watermark method, if made public, then it would be relatively easy to destroy.

They may try to protect the data in the watermark with strong encryption, but since we are (1) not interested in knowing the data, (2) not interested in modifying the data, and (3) all we want is to destroy the watermark...

It should be relatively easy. Just introduce some random noise targeted specifically at the watermark.

And thus to answer spath: We might have to use specialized method instead of generic transforms. But in this case, hackers that are also fluent in audiocoding should be able to cook up something.

After DRM, watermarking

Reply #15 –
Quote from: pepoluan on
[...] but since we are (1) not interested in knowing the data, (2) not interested in modifying the data, and (3) all we want is to destroy the watermark...
It should be relatively easy.

Not if the embedding is key-dependent and includes redundancies!

If I encrypt something with AES you can't read my plaintext message just because you know how AES works. You need the key to do it.

The way redundancy is introduced may and possibly should be key-dependent because then an attacker wouldn't know how to destroy the watermark.

my 2 cents,
SG

After DRM, watermarking

Reply #16 –
Quote from: SebastianG on
Quote from: pepoluan on

[...] but since we are (1) not interested in knowing the data, (2) not interested in modifying the data, and (3) all we want is to destroy the watermark...
It should be relatively easy.
Not if the embedding is key-dependent and includes redundancies!

If I encrypt something with AES you can't read my plaintext message just because you know how AES works. You need the key to do it.

The way redundancy is introduced may and possibly should be key-dependent because then an attacker wouldn't know how to destroy the watermark.
Just because you are the only one with your AES key doesn't mean someone cannot destroy the file so that you can no longer get your plaintext message out or even identify that it was indeed your file to start with.  Your analogy doesn't hold because an AES-encrypted plaintext file serves no other general purpose than hiding what's inside.  The point here is that the audio is available without a key, and only the watermark is supposedly hidden.  If you don't care to find the key but just want to destroy it so that you cannot be tracked, that is potentially quite easy.

After DRM, watermarking

Reply #17 –
Quote from: rsdio on
Just because you are the only one with your AES key doesn't mean someone cannot destroy the file so that you can no longer get your plaintext message out or even identify that it was indeed your file to start with.  Your analogy doesn't hold because an AES-encrypted plaintext file serves no other general purpose than hiding what's inside.
Sebastian's analogy is perfectly valid, you just didn't get it.
Security algorithms whose strength depend only on the key
and not on the secrecy of the algorithm are said to satisfy
Kerckhoffs' principle ; cryto algorithms like AES satisfy it,
as well as secret-key watermarking algorithms.
Quote from: rsdio on
The point here is that the audio is available without a key, and only the watermark is supposedly hidden.  If you don't care to find the key but just want to destroy it so that you cannot be tracked, that is potentially quite easy.
Everything looks easy when you don't know how it works.

After DRM, watermarking

Reply #18 –
Quote from: spath on
Sebastian's analogy is perfectly valid, you just didn't get it.
Security algorithms whose strength depend only on the key
and not on the secrecy of the algorithm are said to satisfy
Kerckhoffs' principle ; cryto algorithms like AES satisfy it,
as well as secret-key watermarking algorithms.
Quote from: rsdio on

The point here is that the audio is available without a key, and only the watermark is supposedly hidden.  If you don't care to find the key but just want to destroy it so that you cannot be tracked, that is potentially quite easy.
Everything looks easy when you don't know how it works.
I concede that it is not easy to remove a watermark that you cannot identify.  Labels will likely use secrecy to prevent easy circumvention.
However, I still believe that the analogy is far from perfect.  An AES-encrypted plaintext message serves no purpose without the key.  I can easily "destroy" it such that you cannot get your plaintext out again, even if you have the key.  For example, I can encrypt your AES-encrypted payload within my own encrypted container.  However, I can do this only because your encrypted message doesn't play a pretty song for me that I wish to preserve.  The real problem is that watermarks are potentially encrypted payloads within unencrypted content.  Thus, the task of destroying a watermark in a song is vastly different than the task of destroying a simple AES-encrypted message.  I simply take objection to using an inappropriate analogy.
Perhaps we're saying the same thing here.

After DRM, watermarking

Reply #19 –
Quote from: SebastianG on
Quote from: pepoluan on

[...] but since we are (1) not interested in knowing the data, (2) not interested in modifying the data, and (3) all we want is to destroy the watermark...
It should be relatively easy.

Not if the embedding is key-dependent and includes redundancies!

If I encrypt something with AES you can't read my plaintext message just because you know how AES works. You need the key to do it.

The way redundancy is introduced may and possibly should be key-dependent because then an attacker wouldn't know how to destroy the watermark.

my 2 cents,
SG



Suppose 50 people who all bought differently watermarked copies all get together and calculate 1/50 (sum of each sample) as a file?
-----
J. D. (jj) Johnston

After DRM, watermarking

Reply #20 –
Quote from: Woodinville on
Suppose 50 people who all bought differently watermarked copies all get together and calculate 1/50 (sum of each sample) as a file?

Then we might be able to identify up to 50 leakers, because for the same key-dependent
watermarking algorithm you don't know if the watermarks coming from different keys will
interfere at all.

After DRM, watermarking

Reply #21 –
Quote from: spath on
Quote from: Woodinville on

Suppose 50 people who all bought differently watermarked copies all get together and calculate 1/50 (sum of each sample) as a file?

Then we might be able to identify up to 50 leakers, because for the same key-dependent
watermarking algorithm you don't know if the watermarks coming from different keys will
interfere at all.



Even if we have key-dependent methods, how would you prevent cross-interference at the modem level?

Yes, I do know how to use the "keys" in a modem. You'd have to have the keys using very different parts of the signal in order to avoid substantial interference.
-----
J. D. (jj) Johnston

After DRM, watermarking

Reply #22 –
Quote from: spath on
Quote from: Woodinville on

Suppose 50 people who all bought differently watermarked copies all get together and calculate 1/50 (sum of each sample) as a file?

Then we might be able to identify up to 50 leakers, because for the same key-dependent
watermarking algorithm you don't know if the watermarks coming from different keys will
interfere at all.

  Yes but with 50 copies, you can identify the ectropy, which is probably very close from the original and without a coherent watermark...
The entropy represents the watermark tho.

After DRM, watermarking

Reply #23 –
Quote from: Woodinville on
Suppose 50 people who all bought differently watermarked copies all get together and calculate 1/50 (sum of each sample) as a file?

That's certainly a valid attack i didn't think of. I was focusing on this scenario: Attacker gets one audio file watermarked with key k and knows only the watermark algorithm (not the key). I believe it's hard to destroy the watermark (given a good "key-dependent algorithm") without severe quality loss.

I'm not sure how well current watermark schemes can cope with averaging multiple (count: n) different watermarked audio files (different keys / same keys). It should be possible to detect all watermarks up to a certain n if different "keys" have been used -- bounded by redundancy and whatnot. That's of course assuming that the watermark signals don't interfere (i.e. orthogonal). Whether this is a reasonable assumption for watermark signals produced with the same "key" I don't know ... (possibly not)

/me only knows the basic crypto stuff which is taught at universities and doesn't include watermarking.

Cheers!
SG

After DRM, watermarking

Reply #24 –
Quote from: SebastianG on
Quote from: Woodinville on

Suppose 50 people who all bought differently watermarked copies all get together and calculate 1/50 (sum of each sample) as a file?

That's certainly a valid attack i didn't think of. I was focusing on this scenario: Attacker gets one audio file watermarked with key k and knows only the watermark algorithm (not the key). I believe it's hard to destroy the watermark (given a good "key-dependent algorithm") without severe quality loss.

I'm not sure how well current watermark schemes can cope with averaging multiple (count: n) different watermarked audio files (different keys / same keys). It should be possible to detect all watermarks up to a certain n if different "keys" have been used -- bounded by redundancy and whatnot. That's of course assuming that the watermark signals don't interfere (i.e. orthogonal). Whether this is a reasonable assumption for watermark signals produced with the same "key" I don't know ... (possibly not)

/me only knows the basic crypto stuff which is taught at universities and doesn't include watermarking.

Cheers!
SG


Well, it seems to me that 'n' is the only question here. Enough sums and you'll have diluted each watermark signal, even if they are orthogonal.

An interesting question is "will this affect how the music sounds".
-----
J. D. (jj) Johnston