Researchers at the Fraunhofer Integrated Publication and Information Systems Institute have successfully tested a software system, based on the group's own digital watermarking technology, for tracking pirated audio files in P-to-P (peer-to-peer) file-sharing networks, said Michael Kip, a spokesman for the institute.

...

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 both sound and image files.

...

The digital media watermark used in the Fraunhofer system also contains a "hash value," which creates a link between the content provider and registered purchaser. "The hash value is like a fingerprint; it contains unique information about the user," Kip said. "The software that we've developed can automatically search for fingerprints."

The Fraunhofer approach differs from others in that it doesn't monitor the individuals who illegally download music but rather scans for content that has been illegally uploaded.

"If, for instance, you purchase and download a CD, burn a copy and give it to a friend and that person puts it on a file sharing network, our system will trace that music back to you and, depending on the legal system of the country you're in, you could be [hit] with an expensive fine," Kip said. "This could certainly help deter online music piracy."

Fraunhofer envisions the prototype software as an application that content providers can install on their own servers for automatically monitoring P-to-P networks around the clock.

What is the Watermark

The Watermark of ISPI Fraunhofer Institute ? An efficient technology to prevent copyright infringement

Each sold track is provided with this special watermark to enable tracking of its source.
It includes the name, adress, personal registration number and a secred code - all saved in the track.
The explicit watermark will be permanently embedded in the file, even after it is burned onto CD or converted to other file formats.

Can I share my downloads with friends?

No. Anyone found distributing, sharing or copying our MP3 files will have their account suspended and face prosecution by the relevant authorities.
Piracy will not be tolerated.

either:
2) burn as audio cd and retranscode as unwatermarked mp3 (meaning even more worse sounding, second generation mp3s)

*  this is only relevant if the record industry decides to allow online music stores to sell unprotected audio files.

Apparently, this is designed to survive such a process.

I doubt that's possible.  If it is, they could just put this technology on CD's in the first place.  Personally, I think they're just trying to scare people.

Putting it on CD's means deploying lots of different copies (different data) of the same CD, which means increasing the costs of production quite much.

Why different copies/data?
I can imagine studios simply running their finished product through a 'watermarker'.
If it survives transcoding and stuff it might be possible to apply it to the raw pcm data too.

although, I hope that this won't happen, I don't like the thought of having more or less personal data stored in my audio files.

watermark = digital print.

To have a different watermark for each CD, it means that the data of each cd is different.
CD's are not burned. they are pressed. This would require to change that, being each pressed cd unique (as if it was a new title).That's what i wanted to point out.

Being it personal (as in each cd containing the info of the person who buys it) is a much more problematic thing (then, the cd is burned for you, so... shipment requirements, wait x days... etc.). Nothing an user would see as beneficial.

If this was used on CD's would it not also make the Accuraterip database useless as the CRC's would be different for every single CD as the data on each would be unique?

madorangepanda, AFAIK, audio CDs have no CRC data. Adding CRC to the data on audio CD and wrapping it in an usable container makes a WAV or AIFF.

As for the watermarks, from what I've read here, those would not survive the first mp3 encoding, because mp3 removes inaudible data, which means including the watermark (especially Fraunhofer - these codecs remove even audible data  ). The only solution to this would be adding the watermark after encoding to mp3.

And lastly, now that I see other people's opinions about this and think a little more about it, I can't elude the thought - I smell spyware... Imagine Lavasoft Ad-Aware or Zonealarm erasing all your Fraunhofer mp3s (not that I would complain...)

If the watermark is really inaudible, that would be a convenient solution, way better than copy/play-protected files. Unfortunately psychoacoustic models using in lossy encoders and transparent watermark are competiting between each other, and if the watermark is really transparent and the end lossy encoders will destroy it.

isn't it possible to get 2 versions of the same watermarked song and compare the differences between the two. Everything that is present in file 1 but not in file 2 must be the watermark and vice versa. Just a quick thought...

I believe it's possible to embedd a transparent watermak which survives some transcoding passes. Another factor that comes into play is the bitrate of the original "watermark signal" which will be embedded possibly using heavy error correcting codes (low bitrate + high reduncancy). I've read somewhere that FhG is able to safely embedd a watermark signal with a bitrate of around 100 bits/s (no KILO!). Quite believable IMHO.

"If, for instance, you purchase and download a CD, burn a copy and give it to a friend and that person puts it on a file sharing network, our system will trace that music back to you and, depending on the legal system of the country you're in, you could be [hit] with an expensive fine," Kip says. "This could certainly help deter online music piracy."

It isn't clear if the burned CD is an audio CD...

Watermarking (or better the identification by a watermark) requires a bitperfect copy of the file.

Watermarking (or better the identification by a watermark) requires a bitperfect copy of the file. It's very unlikely that a transcoded file retains a 100% undisturbed watermark.

Hm, could you elaborate a bit more, what I have to understand under "relative features of the transport"?

...technologies are to be found in research prototypes which are currently at an advanced stage of development.
Feb 9, 2006

Error correcting codes are helpful only if a small percentage of the bitstream is affected.

However, transcoding results in a bit pattern that is randomly different from the original and thus unlikely to be reverted into the original watermark.

Looks like you're thinking "too digital". Good watermarks can be even detected if you decode your file and transmit it via FM radio. This is already done by some stations and gadgets exist that record 10 seconds of such a signal and tell you exactly which artist is singing what title -- 'cause that's what has been hidden in the signal (watermark).

Sebi

Depends on the redundancy which can be very high and probably is.


Looks like you're thinking "too digital". Good watermarks can be even detected if you decode your file and transmit it via FM radio. This is already done by some stations and gadgets exist that record 10 seconds of such a signal and tell you exactly which artist is singing what title -- 'cause that's what has been hidden in the signal (watermark).

Sebi