Using Watermarks to Combat Piracy

TheEvilOverlord writes to tell us PC Advisor is reporting that researchers at the Fraunhofer Integrated Publication and Information Systems Institute have developed a new watermarking system to help track and combat piracy. From the article: "The system lets content providers, such as music studios, embed a watermark in their downloadable MP3 files. Watermark technology makes slight changes to data in sound and image files. For instance, the change could be a higher volume intensity in a tiny part of a song or a brighter colour in a minuscule part of a picture. Even the best-trained human eyes and ears, according to Kip, can't detect the change."

Comments

Human?

[biocute]

Even the best-trained human eyes and ears, according to Kip, can't detect the change.

Who says anything about using human senses to detect the watermark? If these watermarks are embedded by machine, I'm sure it won't be long until Watermark Bob creates a "cleanser" program to detect anything unusual, and maybe even remove it.

Re:Human?

[Iphtashu Fitz]

I'm sure it won't be long until Watermark Bob creates a "cleanser" program to detect anything unusual, and maybe even remove it.

Good point. All you'd really need is two or more copies of a given file, each with their own watermarks. Do a relatively straightfoward binary diff on the files and you'd quickly spot the watermarks. Normalize the diffs based on the similarities between the multiple file copies, and voila! Instant un-watermarked file.

Re:Human?

[dustmite]

That's a valid intput, but steganographers thought of that years ago already. Decent steganographic techniques include low-frequency information that can make them quite resilient to a fair deal of subsampling, recompressing, re-encoding and so on. The idea is not to make a "miniscule variation" but a very subtle variation over a large area. You can think of it like, the actual information is in the 'high bits' not the 'low bits'. Info in the 'low bits' is easily destroyed.

Re:Human?

[dorkygeek]

Huh? I guess you don't understand. Every legal copy gets a different watermark, and the buyer is registered. If somebody thinks you have an illegaly copied file, they can trace back to the original buyer, who spread the file.

Re:Human?

[LocoMan]

3dbuzz.com does that on the maya training videos they sell, basically when you buy it, the copy that will be shipped to you gets encoded with a near invisible watermark with your name, address and phone number (or at least the credit card owner's name, address and phone number), and they're a much smaller operation than the ones this refers to, so I guess that with enough computer power watermaking a video on the fly would be not only posible, but practical as well. As parent said, though, how long the watermak will last is another matter.

Re:Human?

[sabernet]

Instant defense: Sony's rootkit allowed a hacker to hijack my PC and steal my files;)

Re:Human?

[Kadin2048]

The intended use of watermarking, at least as I see it, is less for mass-released files than for prerelease or limited-release uses.

E.g., radio stations get copies of songs weeks before the CDs actually hit stores -- and suspiciously, the songs tend to show up on P2P networks soon after they go out to radio stations. What the music producers would really like to be able to do is trace the leaked files back to whoever put them on the internet, and then get medeival on them for breach of contract.

You can imagine similar uses for prerelease screenings of movies that go out to critics, film review boards, etc. It's less about preventing piracy than it is about finding the snitch afterwards so they can be made an example of. Really, the piracy deterrent is not technological (the watermark), but social (whatever punishment gets inflicted). The watermark is just facilitating the latter.

I suppose in theory if you had a watermark that could be embedded into the file quickly and easily, you could use it on downloaded music (like the iTMS) to see if people were sharing files that they purchased, but really I think systems like this are designed to catch big fish, not Joe Preteen who's ripping files that he bought off of Napster and putting them onto Kazaa.

A lot of similar systems are used with images; actually many of the techniques used for watermarking are used for steganography (it's arguable that watermarking is really a form of steganography), like Least Significant Bit padding for one. There are also systems that have a robust enough watermark that they will survive printing and scanning, although they tend to begin to mess up the image slightly.

Re:Human?

[pclminion]

Who says anything about using human senses to detect the watermark? If these watermarks are embedded by machine, I'm sure it won't be long until Watermark Bob creates a "cleanser" program to detect anything unusual, and maybe even remove it.

The purpose of making the watermark imperceptible isn't to trick the user. The user is fully aware that the music is watermarked. The reason it is done the way that it is, is because it is the only way to do it. There are really two requirements for the watermark. First, it must be intertwined with the actual data in order to make it non-trivial to remove. Simply sticking the watermark in a meta-information block would make removing it too simple. Second, because the watermark is intermixed with the audio data, it MUST be done in an imperceptible way in order to retain the quality of the recording.

So, making the watermark inaudible is not some attempt to pull one over on you. It's just the only realistic way it could be done.

Um, what?

[Jugalator]

The system lets content providers, such as music studios, embed a watermark in their downloadable MP3 files

For whom was this intended again?

I'd be happy if there actually was plenty of music studios providing downloadable mp3's though.

Defeating?

[Anonymous Coward]

...and in order to defeat such a wonderful scheme, all you have to do is re-watermark the image/music/video.

I've yet to see a scheme that reliably survived that test unless it was specifically designed just for that test (like embedding high power signal in several random places), and upon detection, looking for that signal in those random places (hope is that 2nd watermarking didn't wipe out -some- signal data).

In any case, Watermarking doesn't work! Even Microsoft's researchers said so (damn, can't find link).

Re:Stand up to Encoding?

[Jordan Catalano]

"How well will this stand up to a lower bitrate/encoding setting?"

About as well as my ears do, I'm guessing.

Nothing to see/hear

[Tackhead]

> The system lets content providers, such as music studios, embed a watermark in their downloadable MP3 files. [ ... ] Even the best-trained human eyes and ears, according to Kip, can't detect the change.

In other words, "Nothing to see/hear. Please move along?"

More seriously - although it could be stripped out (relatively) easily, you could embed watermarking data in the metadata segments of downloadable MP3s. I'd accept this as a tradeoff for music studios offering downloadable MP3 files: If some_hit_song_i_downloaded.mp3 shows up on a P2P network and contains metadata whose MD5 could only be generated by, say, hashing my credit card number with some_riaa_private_key, that'd be pretty reasonable grounds for RIAA to believe that I'm the schmuck who (a) paid for the right to download it from a RIAA-authorized source, and (b) uploaded it to a non-RIAA-authorized filesharing network.

Make it impractical for Joe Sixpack (who will be unaware of this type of watermarking, and who probably will be unaware of the existence of tools to strip it) to upload his files without risking fines/prosecution, and you can offer DRM-free MP3s to Joe Sixpack.

Re:Nothing to see/hear

[AviLazar]

Since your above example utilizes credit card information you CANNOT make Joe Sixpack unaware. He has to know, explicitly, that his credit card information is being encoded and sent up. Joe Sixpack needs to be given the absolute right to say "No I really do not trust your security system, and I will not give you my GC #"

Re:And this fights piracy how?

[Iphtashu Fitz]

OK, so you tag downloads. Now what?

In theory it lets the distributor figure out who the source of the piracy was. Joe User logs into their site and downloads the latest hit DRM_SUX.mpeg. Unknown to him it has a unique watermark in it that identifies him as the one who downloaded this particular file. Six months later the Copyright Kops find a copy of DRM_SUX.mpeg floating around on P2P networks. They analyze the file and discover the watermark points to Joe User, so they then sick their landsharks^M^M^M^M lawyers on him.

Re:And this fights piracy how?

[Billosaur]

Assuming a "de-tag" program doesn't pop up an hour later, what do you do with this wonderful invention? Instead of passing around a "normal" mp3 of Metallica, they're now sharing a "watermarked" version that allegedly can't be discerned by mere humans. How does this help?

You code media players to detect the watermark (which would have to be in a standardized format) and refuse to play anything that does not contain the watermark. Conversely, ripping programs will not rip anything containing the watermark, making it harder to copy the source. You wouldn't have to worry so much about removal programs, as programs that would "fake" the watermark, basically couterfeiting programs. Of course, those would pop up fifteen minutes later.

That's the spirit

[RyoShin]

I'm hoping these kinds of anti-piracy actions work, and work well.

Things like the DRM and DMCA were put into place to fight piracy, and wound up just hurting regular consumers while the pirates just snickered as they continued along their merry way.

With these kinds of things, regular users will still be able to do what they like with their own copy, be it back it up or transfer it to another medium for personal use. At the same time, it will allow those tracking piracy to find the source and press charges only against that person, and not the random multitude.

I'm sure the pirates will figure out some way to work around this (be it to randomly change the volume slightly throughout an entire MP3, or brightening/dithering an entire picture), as they have everything else, but if this kind of technology can prevail and advance, it will allow those of us legally using our own purchased goods to do so without worry, while punishing those who deserve it.

Re:That's the spirit

[no_opinion]

If the majors sold inaudibly watermarked but non-DRMed MP3 files, would people buy them? My friends say no because they think people will find this too "big brotherish" but I think that the only ones at risk are the people violating copyright (i.e. sharing on p2ps). I'm willing to buy an INAUDIBLY watermarked mp3 file, because then I can do whatever I want with it, I don't have to worry about DRM, and I'm not at risk because I'm not infringing copyright.

What do others think?

Re:That's the spirit

[markdavis]

It scares me. You could then be held liable, if someone broke into your system and "stole" some of the watermarked works and then proceeded to "share" them with others. Even worse, someone could frame someone else by simply re-watermarking media to have it point to someone else.

Although it sounds like a very attractive alternative to DRM, there are some serious security issues that would have to be adequately addressed.

Key question

[overshoot]

However, the question is how this gets applied.

Are they planning to

• ship millions of identical watermarked files, then expect hardware to refuse to play any that files that aren't licensed (same old business model, someone else's problem.)
• Mark each file to identify the purchaser, then go after the source of widespread copyright violation?

The first is basically worse than DRM, the second is essentially an aid to enforcing existing copyright laws. I suspect that if the Content Cartel would finally accept that their business models need to change and go for the second approach, most of us could accept it.

Short memories -- this was called SDMI

[Thagg]

This was rolled out years ago, and plotzed with a mighty thud when it happened, due in no small part to the http://www.cs.princeton.edu/sip/sdmi/faq.html">wor k of Felten and his grad students at Princeton.

Basically, the Powers That Be came up with a very good watermarking system, but even the best system can be defeated by a very determined adversary -- especially since the watermarks can't be updated once the CDs are shipped.

Another problem that I've always had with these systems is the proof issue. If the RIAA tries to prosecute you for having watermarked files, they have to demonstrate the watermark. I can't imagine how they could show that without revealing exactly how the watermark is detected -- and once they do that, you should be off to the races.

Anyway -- this has been tried, and it has failed. The SDMI system was really quite sophisticated, and it failed almost immediately.

Thad Beier

Re:Short memories -- this was called SDMI

[Fnord666]

Another problem that I've always had with these systems is the proof issue. If the RIAA tries to prosecute you for having watermarked files, they have to demonstrate the watermark. I can't imagine how they could show that without revealing exactly how the watermark is detected -- and once they do that, you should be off to the races.

This is exactly what "zero knowledge" proofs are supposed to be able to do. Using such a protocol allows me to prove to anyone interested that I know something without revealing what that something is. the proof does not impart any new knowledge to the person/entity that is receiving it about the information. In this context it would mean that the RIAA would be able to prove that the files had been watermarked using their private mark and were thus their copyrighted material without revealing any information about how that watermarking was done. The proof would not rely on revealing the method. Graph isomorphism is a commonly used example.

I don't know if anyone has incorporated a practical application of this into any sort of watermarking scheme, but it could be used for this purpose.

This would actually be great

[grahamsz]

MP3's such a universally accepted format that i'd be able to purchase music online and be able to use it wherever i want - be it in the gym, on my ipod, on the tivo, and mac/pc/linux.

Watermarked MP3s would be a way that the music industry could say "look, we almost trust you!"

This will not replace current "DRM" schemes

[massysett]

because it puts the enforcement burden on the record labels. There would be millions of watermarked files floating around out there, and they'd have to sue enough people to scare folks into not sharing their files. Only through scaring people could the labels have enough impact with this--there's no way they could close the spigot with the lawsuits.

Even so, the labels might adopt something like this. But it would be in addition to their current copy restriction schemes, rather than a replacement for them. Consumers still lose as they'll still have to wrangle with FairPlay, WMA, or whatever copy restriction scheme the labels want to use.

I don't understand....

[mangu]

...how do they mark the water, and how do such marks on the water keep pirate ships from attacking honest merchant ships?

Re:Why should Clear Channel care about file sharer

[sam_handelman]

I respectfully disagree.

  Your problem is that you are accepting the recording industry's propaganda, i.e. "We oppose piracy because people will listen to pirated copies instead of buying CDs."

  The *real* objection of the recording industry, and this goes double for clear-channel, is that P2P sidesteps their promotion monopolies and makes the music market harder to manage and control. Fragmentation of the market costs them their niche at the top of the foodchain.

  The best example of this attitude was, a while back, movie industry executives noticed that some heavily promoted presumed-blockbuster (I forget which movie it was, The Island maybe) was getting far less than the guaranteed level of attendance given the advertising budget. Careful marketing research traced this phenomenon back to bad word of mouth, which was spreading faster than it had in the past, chiefly by cellphone.

  The response of the movie industry was NOT "gee, we'd better stop making movies that even brain damaged 11 year-olds regard as intellectually insulting", but instead "is there any way we can make it illegal to badmouth our movies by text message? Libel law, maybe?" Fortunately, they concluded that was a non-starter.

  That long tangent aside, look at clearchannel. Clearchannel's business model depends COMPLETELY on the willingness of the general public to agree-to-like whatever 30 songs they decide they want to play/promote in a single month. They also need to make sure that people keep listening to the radio and not to ipods. Alternate routes of distribution are just as much a threat to clearchannel as they are to the recording industry.

Sounds great (not).

[marcello_dl]

Also it requires every purchaser of a copy to be a registered one, it requires the purchaser to be very careful not to have the copy stolen or lose it, and it might also lead a hacked watermark to accuse an innocent purchaser.