Generating a clean digital copy using analogue transmission

[Shalmanese]

The whole uproar about file sharing as opposed to lending tapes to friends is that with digital music, its possible to make bit-for-bit copies of the data without any degradation of quality. However, it occurs to me that instead of making ONE analogue copy, someone could make MULTIPLE analogue copies of a disk and then compare each copy with the others. If the bit error rate is above 50%, then over enough copies, the "true" value of the bit can be determined with enough precision. Furthermore, I guess if you play a known good track over the same line, you could also scrub most of the systematic noise in the system as well. In the end, you should end up with a fairly good copy of the original, furthermore, that copy is in digital form and without DRM so you can copy it to somewhere else without any hassles.

Does this mean that DRM is effectively dead in the water since theres no way you can really ever prevent someone from making analogue copies? Has anybody tried doing this before?

 

[rjain]

Er, that's how digital information is stored and transmitted anyway. All media are analog, it's just the interpretation of certain levels as being a 1 and certain ones being a 0. Then you have ECC/parity/CRC to help deal with signal degradation. Much easier to do it that way than to have multiple copies of the same data (Although that is clearly the more reliable, but doesn't offer that much benefit for the extra capacity you're wasting. You'd do better to use a more powerful ECC scheme if you have the capacity to waste.)

 

[Matthias99]

I think what he's suggesting is the following way to get around DRM:

Take a clean, encrypted digital copy with DRM.

Play it back multiple times, making a new, unlocked digital encoding (presumably in an uncompressed lossless form like CDA) of the analog output from your soundcard.

Compare the new (unencrypted) digital copies, and try to determine if there are any errors in the new recording relative to the old one by seeing if they disagree on any bits.

While this might work, the fact remains that you're losing some fidelity by transferring the data from digital to analog and back (with most being lost in the analog-analog transfer from the speaker output to the recording input, and a little bit in the analog->digital step at the end). There are two problems I see. First, it's very unlikely that encodings of two analog copies of a song would line up bit-for-bit. Even a few milliseconds time difference between the samples will make a comparison like this difficult, if not impossible. You may be able to eliminate a lot of line noise, though. Second, most of the degradation is going to come from whatever mic/recording input you use. Sampling theory says that you can't increase the amount of data present in the recording once it's been encoded -- to get higher quality, you'd have to get better equipment that takes away less of the analog data in the first place.

People have done this (and will continue to do so, I'm sure), although I'm not sure anyone has tried the multiple copies thing. However, any DRM system can be broken digitally-- obviously the player has a mechanism to decode the data, so all you have to do is figure out how that mechanism works (which is not that difficult; they usually publish papers on this sort of thing) and then spend sufficent time to crack its encryption. I'm sure eventually they'll move to systems with impractically large (1024-bit or so) keys eventually, but there are still other ways around it (such as disassembling the playback executable and trying to rebuild the encryption keys that way, or getting an insider to leak the information, as happened with CSS).

 

[rjain]

Unless you need special sound cards to play DRM-enabled songs, you can just capture the sound (digitally) from the sound card itself, too. If it has a digital output, you can use that, as well.

 

[Matthias99]

Most DRM playback programs won't let you output a protected stream to an unsecure digital output for that very reason. Future versions of Windows are supposed to have similar protections built in for drivers, etc. so that you can't capture digital audio from the sound card output when playing things protected by DRM technology. Isn't the DMCA wonderful?

 

[rjain]

So will it be illegal to write DRM-enabled players for systems that don't have such drivers?

 

[Peter]

Right. Audigy 2 already has the provisions to keep you from doing that kind of thing. Welcome to the free world where the corporations make the laws!

 

[Mark R]

DVD audio is supposed to have features designed to subvert analogue copying as well as digital. Delays in implementing and standardising this system have meant that DVD-A has not yet been supported by labels.

Essentially, the recording on the disc would be 'watermarked' - a barely audible but machine readable sound would be added to the recording. A device playing back the disc would verify that the watermark in the audio corresponds to the security key embedded in the disc.

Detection of the watermark would mean that the soundtrack is copyright protected and the player would transmit a 'do not copy' signal in the digital output. Similarly, if a recorder detects a watermark in an incoming analogue signal, it would refuse to record. If you manage to record a DVD via an unprotected source, the during playback the watermark would be detected and as there would be no matching security key, playback would be blocked.

 

[Shalmanese]

Hmm... what about if you get multiple analogue copies from MULTIPLE disks? that way, you would blur the water mark out.

Matthias: Are you saying that analogue copies actually have time jitter as well? I did not know that and I could see how that would affect the process. If theres not time jitter, you could easily synch the two copies by playing a reference wave before and after the piece is done to let the software lock onto to the signal.

 

[CTho9305]

Originally posted by: Shalmanese
Hmm... what about if you get multiple analogue copies from MULTIPLE disks? that way, you would blur the water mark out.

Matthias: Are you saying that analogue copies actually have time jitter as well? I did not know that and I could see how that would affect the process. If theres not time jitter, you could easily synch the two copies by playing a reference wave before and after the piece is done to let the software lock onto to the signal.

Or just pass it through a low-pass filter at about 20khz and a high-pass filter at about 20hz, and any inaudible watermarks would be gone.

 

[Mark R]

Matthias: Are you saying that analogue copies actually have time jitter as well? I did not know that and I could see how that would affect the process. If theres not time jitter, you could easily synch the two copies by playing a reference wave before and after the piece is done to let the software lock onto to the signal.

This is a basic problem with trying to combine multiple analogue samples. In order to get bit for bit accuracy, you would need to ensure that the DAC and ADC are synchronised and remain stable at within 15 ps. This is essentially not achieveable unless the 2 converters run from the same clock.

Hmm... what about if you get multiple analogue copies from MULTIPLE disks? that way, you would blur the water mark out.

Well, I would assume that the watermark would be the same on every copy of the disc. So unless you got the single version, album version and some compilation re-release this wouldn't work - and then you have the added problem of making sure that the combined versions are infact the same.

Or just pass it through a low-pass filter at about 20khz and a high-pass filter at about 20hz, and any inaudible watermarks would be gone

Not quite - the watermarking is in the audible band so simple filtering will not affect it. It is added using psychoacoustic principles with emphasis on inaudible bands (e.g. MP3 compresses so effectively by working out which sounds you can't hear - the watermark is specifically added to those places). It is more sophisticated than that however, because the watermarks are designed to withstand MP3 encoding - I understand that current versions will survive MP3 encoding to 96 kbps.

 

[NeoPTLD]

Modem can transmit and receive digital signal w/o loss of information. if you can include CRC/Parity into the analog signal like a modem, then you can generate a perfect digital copy using analog transmission.

Originally posted by: Shalmanese
The whole uproar about file sharing as opposed to lending tapes to friends is that with digital music, its possible to make bit-for-bit copies of the data without any degradation of quality. However, it occurs to me that instead of making ONE analogue copy, someone could make MULTIPLE analogue copies of a disk and then compare each copy with the others. If the bit error rate is above 50%, then over enough copies, the "true" value of the bit can be determined with enough precision. Furthermore, I guess if you play a known good track over the same line, you could also scrub most of the systematic noise in the system as well. In the end, you should end up with a fairly good copy of the original, furthermore, that copy is in digital form and without DRM so you can copy it to somewhere else without any hassles.

Does this mean that DRM is effectively dead in the water since theres no way you can really ever prevent someone from making analogue copies? Has anybody tried doing this before?

 

[Shalmanese]

Yes, but under DRM, you can't exactly ask the player to insert CRC information into the stream . Mark: When you were talking about watermarking, I was thinking something akin to the red dots currently in cinemas so that people could track who was doing the ripping.