About Audio Digital Watermarking

[ChunYinZi]

Guys.

Do you know about

Digital audio watermarking?

I understand that some of the library music sites are now incorporating these watermarks.

Digital watermarks are embedded into audio files (wav, mp3, avi, etc.) using watermark embedding algorithms, but they don't affect the sound quality of the audio file much, or are not noticeable to the human ear.

In other words, even if you download a WAV file, the sound quality may not be lossless.

So my advice is that WAV on CD is the best.

What do you think?

[Psyclon]

You would be surprised how much hidden data can be found and introduced with computer files.
For example, even the "red book" CD - the normal audio CD - has the ability to store CD text. This has been added in the mid-90s mostly from SONY. "CD text" allows to store a few kBytes of data in subchannels on a CD usually for tagging (artist, title, et cetera) so your newly-bought 1998 car stereo would display the title of your current music, similar to your radio stations suddenly having names on your stereo happily displaying messages. The lead-in of a CD for example can contain up to about 5 KBytes of data and may not exceed 3,024 characters. However, 3,000 characters is technically enough to add copyright information - watermarks!

So, these information might be stored on your disk:

[Taken from the Wikipedia]

This is very similar to the tags in your FLAC, AAC, OGG, MP3...file. And here is the DRM and privacy problematic: This is an AAC file of my collection:

ITUNESOWNER and ITUNESACCOUNT contains sensible data - OWNER has my REAL NAME I use for my iPod/iTunes/Apple account (!!!!). OWNER has the eMail address used for that account written there. So if I give you my AAC file in that way, you have my name, my mail address that is obviously valid as it need to be in order to purchase on Apple. If your PC is full of random names, you shared the music. Prosecutors will enjoy reading your confiscated harddisks with all these information in plain sight.

That is why, when I receive files, I clean all that stuff off. Yes, also the data you share (the WAV files) contain dozens of these information. My MP3s are rewritten anyways in order to be encoded and I also manually clean it from all data junk:



When I receive a WAV or FLAC and encode, I clear the meta data in Audacity upon exporting:

But to be honest, with all the sneaky ways of adding tracing and watermarking, this is just the tip of the iceberg. Who knows if even rewritten files contain smears of copyright data...
The meta data that is incorporated are, as far as I know, not embedded into the WAV (the "payload") but attached to it, which would make it still lossless. You must always think of backwards compatibility. If there is hidden information going on in a file, programs might hiccup reading it as it tries to identify that meta data watermark as audio information..

[Fuzi]

Always great to read your analysis, Dr. Psy
🩻

[ChunYinZi]

You would be surprised how much hidden data can be found and introduced with computer files.
For example, even the "red book" CD - the normal audio CD - has the ability to store CD text. This has been added in the mid-90s mostly from SONY. "CD text" allows to store a few kBytes of data in subchannels on a CD usually for tagging (artist, title, et cetera) so your newly-bought 1998 car stereo would display the title of your current music, similar to your radio stations suddenly having names on your stereo happily displaying messages. The lead-in of a CD for example can contain up to about 5 KBytes of data and may not exceed 3,024 characters. However, 3,000 characters is technically enough to add copyright information - watermarks!

So, these information might be stored on your disk:

[Taken from the Wikipedia]

This is very similar to the tags in your FLAC, AAC, OGG, MP3...file. And here is the DRM and privacy problematic: This is an AAC file of my collection:

ITUNESOWNER and ITUNESACCOUNT contains sensible data - OWNER has my REAL NAME I use for my iPod/iTunes/Apple account (!!!!). OWNER has the eMail address used for that account written there. So if I give you my AAC file in that way, you have my name, my mail address that is obviously valid as it need to be in order to purchase on Apple. If your PC is full of random names, you shared the music. Prosecutors will enjoy reading your confiscated harddisks with all these information in plain sight.

That is why, when I receive files, I clean all that stuff off. Yes, also the data you share (the WAV files) contain dozens of these information. My MP3s are rewritten anyways in order to be encoded and I also manually clean it from all data junk:



When I receive a WAV or FLAC and encode, I clear the meta data in Audacity upon exporting:

But to be honest, with all the sneaky ways of adding tracing and watermarking, this is just the tip of the iceberg. Who knows if even rewritten files contain smears of copyright data...
The meta data that is incorporated are, as far as I know, not embedded into the WAV (the "payload") but attached to it, which would make it still lossless. You must always think of backwards compatibility. If there is hidden information going on in a file, programs might hiccup reading it as it tries to identify that meta data watermark as audio information..

Many thanks to Psyclon for the analysis!

I don't know if you remember the SourceAudio series of websites? The WAVs you download from there are actually digitally watermarked.


I spoke to some of the people in charge of SourceAudio and they told me that any music on their site is digitally watermarked.


Their purpose is to monitor the flow of music, and in fact, they've been monitoring radio and TV shows, and if the system analyzes their digital watermarks, they know what's going on with the music licensing It also makes it easier for them to check for licensing or copyright infringement, and it's just a way to keep track of what they're doing. The details... it's explained on their official website.



As a result, I rarely download music from SourceAudio anymore, I think they are all lossy wavs, but of course that's probably just me being psychotic because the human ear can't hear the digital watermark.


Currently, if I have a particular favorite piece of music, I tend to buy the CD and digitize it, which I think is the best thing to do.

[Psyclon]

I spoke to some of the people in charge of SourceAudio and they told me that any music on their site is digitally watermarked.

Their purpose is to monitor the flow of music, and in fact, they've been monitoring radio and TV shows, and if the system analyzes their digital watermarks, they know what's going on with the music licensing It also makes it easier for them to check for licensing or copyright infringement, and it's just a way to keep track of what they're doing. The details... it's explained on their official website.

I'd like to see that explanation. To me, it sounds more like "scaring off people" that are not very tech-savvy, almost like the famous "Don't copy that floppy" campain in the 90s or the various "If you burn that CD, you are a criminal" messages which are often plainly wrong or just scary bedtime stories from the music industry that are again, wrong, or never enforced by, well, being wrong.

I am almost completely sure you can strip a WAV from any filthy attached metadata and the musical data, the aformentioned payload, COULD technically contain signals in certain ranges (e.g. 20-22 KHz) that translate into information and text, almost like morse code and still inaudble for the human ear. But then, you could simply lowpass filter that away to e.g. 18 kHz, and everything below would indeed meddle with the quality of the sound of the actual music, degrading the quality so much that nobody would want to license that music anymore. There are so many ways to add signals and data, like RDS (radio data systems), teletext/ceefax, subtitles over analogue TV et cetera, so I would never say never, but much of the stuff seem to be somewhat unrealistic.

It remembers me so much to a bus driver here that has a dashcam and he told some older lady where there is a camera about all kinds of nonsense, that this little thing would pixelate faces for anonymity but with a certain master password, the police could unblurr it et cetera while I just smiled away - as if that 50€ thing powered by 12V has the ability to write two video streams simultaneously and add face recognition and such but the lady believed that. Probably to scare off vandals ruining the seats covers et cetera, but for anyone with a little bit of knowledge, that was just a funny story to make your day.

Currently, if I have a particular favorite piece of music, I tend to buy the CD and digitize it, which I think is the best thing to do.

That is always the best. You own the disk yourself and while there is, as said above, very possible that the disk itself has copyright embedded, you have the physical thing. It is also mentionable that owning the music is less of a problem, that's why I think LMT is in a grey area as broadcasting/using/making money is where the problem starts. I know there is the "unauthorized duplicating" thing, but all of this is difficult to actually enforce in our situation here. EDIT: Also to mention that that metadata is often written "beside" the actual PCM stream, not altering the music information in any way to keep it lossless.

[nidostar]

This is fascinating stuff. So are digital watermarks just metadata tags or are they transmitted with the digital audio as a kind of sub signal much as RDS data is broadcast over the radio in real time? I am assuming the latter since ChunYinZi mentions that copyright owners monitor radio and TV stations to see whether the broadcast of their music is properly licensed. Though with the wealth of terrestrial and internet radio stations worldwide that must surely be an impossible task? At least we have control over the vinyl rips that we make.

[Psyclon]

Yup, Nidostar, I also wonder how they monitor that. For example, you have a legit CD copy of my song above, "Mercy" from Steve Jones. That, of course, does not contain any copyright data as that disk is from 1987. How can now a random company like SourceAudio come up and request access to your system if the copy you play is theirs. That is not working that way. At first I thought that SourceAudio might NEED to have the tags in place, else a playback/broadcast software might refuse to accept it (as in: "No passport, no entry to the airplane" sort of thing).

Have you ever heard the signal of your radio controlled clock? Here in Germany it is distributed by the radio transmitter DCF77 with a range of 2,000km. It is literally a very poor audio signal with beeps and blips, but within minutes, your radio clock gets time, date, summer time et cetera. With a well-encoded and clear signal embedded within an audio stream in areas where actual music stops (e.g. 20-24 kHz), you could get tons of information going on on continuous playback. That's why I can only speculate and would like to see a FAQ or something.

[ChunYinZi]

I spoke to some of the people in charge of SourceAudio and they told me that any music on their site is digitally watermarked.

Their purpose is to monitor the flow of music, and in fact, they've been monitoring radio and TV shows, and if the system analyzes their digital watermarks, they know what's going on with the music licensing It also makes it easier for them to check for licensing or copyright infringement, and it's just a way to keep track of what they're doing. The details... it's explained on their official website.

I'd like to see that explanation. To me, it sounds more like "scaring off people" that are not very tech-savvy, almost like the famous "Don't copy that floppy" campain in the 90s or the various "If you burn that CD, you are a criminal" messages which are often plainly wrong or just scary bedtime stories from the music industry that are again, wrong, or never enforced by, well, being wrong.

I am almost completely sure you can strip a WAV from any filthy attached metadata and the musical data, the aformentioned payload, COULD technically contain signals in certain ranges (e.g. 20-22 KHz) that translate into information and text, almost like morse code and still inaudble for the human ear. But then, you could simply lowpass filter that away to e.g. 18 kHz, and everything below would indeed meddle with the quality of the sound of the actual music, degrading the quality so much that nobody would want to license that music anymore. There are so many ways to add signals and data, like RDS (radio data systems), teletext/ceefax, subtitles over analogue TV et cetera, so I would never say never, but much of the stuff seem to be somewhat unrealistic.

It remembers me so much to a bus driver here that has a dashcam and he told some older lady where there is a camera about all kinds of nonsense, that this little thing would pixelate faces for anonymity but with a certain master password, the police could unblurr it et cetera while I just smiled away - as if that 50€ thing powered by 12V has the ability to write two video streams simultaneously and add face recognition and such but the lady believed that. Probably to scare off vandals ruining the seats covers et cetera, but for anyone with a little bit of knowledge, that was just a funny story to make your day.

Currently, if I have a particular favorite piece of music, I tend to buy the CD and digitize it, which I think is the best thing to do.

That is always the best. You own the disk yourself and while there is, as said above, very possible that the disk itself has copyright embedded, you have the physical thing. It is also mentionable that owning the music is less of a problem, that's why I think LMT is in a grey area as broadcasting/using/making money is where the problem starts. I know there is the "unauthorized duplicating" thing, but all of this is difficult to actually enforce in our situation here.

Here's the official digital watermark information from SourceAudio

htXXs://www.sourceaudio.com/detect/

My conclusion is that if you can, try not to download wavs from SourceAudio.


Also, I think Universal Music, APM Music, and others may have added digital watermarks, but there's no official word on it.

So, I've been searching for CDs of my favorite music.

[Psyclon]

Thank you for the link and I checked the Digimarc website too which has a SourceAudio case study report. It is far more sophisticated than I would have thought! It is actually a carrier signal over the actual audio and you can do it by yourself at anytime if you wish - by adding a Digimarc Barcode to any file, even pictures.

A noval data carrier that encodes data in a media in ways that are generally [sic!] imperceptible to people, permitting the carrier to be repeated many times over the surface of the Enhanced media. it delivers unprecedented ease of use, reliablility, and efficiency in identifying the media due to massive methematical and graphicl redundancy.

Yes, it is indeed "interwoven in the fabric of the audio files" and even for me as MP3 "fan", I think that's very disgusting. "Generally imperceptible" means they alter the files and apparently add errors and artifacts to it sometimes, I mean, "Generally", eh? I assumed they just use "unnecessary" bandwidth to store data (20-24 kHz and embedd data there). They also offer this for JPGs and said you can crop it down to about 256x256 pixel and it'd be still intact, so I guess they had the option of lowpass filtering in mind for audio too. But then, they also say thave it interwoven into lossy encodings and that appears to be reliable. 0.2 seconds to detect so many information is strangely fast. I am highly interested of how that really works though, I would not be surprised if there is just a combination of shenanigans going on of different things here, but they are obviously very vague about it. But yes, if they add such a big junk of information directly into the payload, these files are not lossless anymore indeed.

SourceAudio Detect, powered by Digimarc, does this without distorting the original audio file. In fact, tests with longtime pro sound, recording and mastering engineers, all confirmed the code was inaudible.

[Digimarc.com]

This could mean anything. id3 tags (of your FLAC and MP3) do not alter it too as it's not within the payload as the tags are seperated from the actual music data. But they indirectly claim this is embedded in the payload, and by that, it is altering the audio.

In general, I'd stay away from that. However, and here is the big deal, I still wonder how that really works. I do understand that, if you record your own music today and give it to SourceAudio for distribution, they will "brainwash" it with tags and blockchain digital barcodes and whatnot. I do get that. Of course if a random station is playing your song but you received €0 so far, they just took it and you get reports just like Youtube content ID matches. But for existing music? Apparently still fingerprinting which is comparably unreliable. And you have to opt-in.

The monitoring and reporting platform inaudibly embeds a digital code from Digimarc in tracks uploaded to SourceAudio’s cloud-based music database by clients that have opted into the program.

[Digimarc.com]

In the end, it could be everything. I am just intrigued that with 0.2 seconds of playback over radio, they get all the information. That is... just strangely fast in my opinion.

However, when reading about the image watermarking, they admit it can go poof - they call that "survivability":

Image Compression
A digital watermark, in most cases, will survive image compression, but the survival is dependent on
several factors. Lossless compression, such as with PNG, LZW, StuffIt™ and .ZIP formats, does not affect
the survival of a digital watermark because no image data is sacrificed to create the compressed version.
Lossy compression methods such as JPEG or indexed color formats actually remove image data in order to
decrease file size; this can affect a digital watermark’s durability.

So for being "interwoven", it can be actually be made unreliable or completely destroyed. They use an almost invisible "change" to the image in order to plaster their watermark shit all over it, almost like a dog with invisible pee that can be smelled by other dogs...

I do get the importance of watermarking as I read through some forums where musicians said they lose out on so much money. I often thought about my favourite composers making extremely high-quality music, press the CD, comission cover art and maybe two licensees actually pay for it? But man, these people are obsessive and I think you can overdo it. Normally, people want to rip out any kind of DRM out of their pirated files, but if it becomes too obnoxious, people won't enjoy it. Did it help the music industry about CDs? Did it help that BluRays are a pain in the ass to rip even for your legal rights (personal backup copy)? iTunes says you are not allowed to remove DRM, but I did, because I have the legal right of a "platform-independent backup copy", and making it "platform independet" means to rip all "Apple, Inc." out of it. You can even purchase DRM remover legally, but they still added a higher bar, smirking: "Go ahead then", basically denying your legal rights...

I mean, imagine you record a track from the radio and it has this Digimarc jizz all over it as it apparently survives everything like a cockroach. Sorry, f*ck off. You got your rights and money from the broadcast station, so get your s*hit off it, thanks. Would you want it, pirate or legit user or not? I sure would not! Thankfully, my collection is mostly made of CD rips and LP recordings, but the newly added digital releases might have Digimarc all over it. They were apparently already getting active in 2018, but the acceptance of Digimarc wasn't great back then ... I heard they also have ASCAP and BMI and all the right performers on board, so the newly released music might be "infected". In itself not a problem as you don't broadcast, but still, to close up my post:

No, it is NOT lossless anymore!

[nidostar]

Bravo Psyclon. I love it when you get your teeth into things like this. Though sometimes I fear for your health because you become extremely passionate about an individual subject!! 

I remember a similar debate over embedding a carrier signal into videos to prevent pirating. Even going back to the old VHS days which, of course, in retrospect is laughable because the quality of videos was very low res. The same approach though was successfully adopted for DVDs and, as you say, BluRay. I fully understand the need to protect one's copyrighted property but with the advances being made with AI I just wonder how long it will be before any digital fingerprinting can be just as easily removed by the layman.

[ChunYinZi]

Thank you for the link and I checked the Digimarc website too which has a SourceAudio case study report. It is far more sophisticated than I would have thought! It is actually a carrier signal over the actual audio and you can do it by yourself at anytime if you wish - by adding a Digimarc Barcode to any file, even pictures.

A noval data carrier that encodes data in a media in ways that are generally [sic!] imperceptible to people, permitting the carrier to be repeated many times over the surface of the Enhanced media. it delivers unprecedented ease of use, reliablility, and efficiency in identifying the media due to massive methematical and graphicl redundancy.

Yes, it is indeed "interwoven in the fabric of the audio files" and even for me as MP3 "fan", I think that's very disgusting. "Generally imperceptible" means they alter the files and apparently add errors and artifacts to it sometimes, I mean, "Generally", eh? I assumed they just use "unnecessary" bandwidth to store data (20-24 kHz and embedd data there). They also offer this for JPGs and said you can crop it down to about 256x256 pixel and it'd be still intact, so I guess they had the option of lowpass filtering in mind for audio too. But then, they also say thave it interwoven into lossy encodings and that appears to be reliable. 0.2 seconds to detect so many information is strangely fast. I am highly interested of how that really works though, I would not be surprised if there is just a combination of shenanigans going on of different things here, but they are obviously very vague about it. But yes, if they add such a big junk of information directly into the payload, these files are not lossless anymore indeed.

SourceAudio Detect, powered by Digimarc, does this without distorting the original audio file. In fact, tests with longtime pro sound, recording and mastering engineers, all confirmed the code was inaudible.

[Digimarc.com]

This could mean anything. id3 tags (of your FLAC and MP3) do not alter it too as it's not within the payload as the tags are seperated from the actual music data. But they indirectly claim this is embedded in the payload, and by that, it is altering the audio.

In general, I'd stay away from that. However, and here is the big deal, I still wonder how that really works. I do understand that, if you record your own music today and give it to SourceAudio for distribution, they will "brainwash" it with tags and blockchain digital barcodes and whatnot. I do get that. Of course if a random station is playing your song but you received €0 so far, they just took it and you get reports just like Youtube content ID matches. But for existing music? Apparently still fingerprinting which is comparably unreliable. And you have to opt-in.

The monitoring and reporting platform inaudibly embeds a digital code from Digimarc in tracks uploaded to SourceAudio’s cloud-based music database by clients that have opted into the program.

[Digimarc.com]

In the end, it could be everything. I am just intrigued that with 0.2 seconds of playback over radio, they get all the information. That is... just strangely fast in my opinion.

However, when reading about the image watermarking, they admit it can go poof - they call that "survivability":

Image Compression
A digital watermark, in most cases, will survive image compression, but the survival is dependent on
several factors. Lossless compression, such as with PNG, LZW, StuffIt™ and .ZIP formats, does not affect
the survival of a digital watermark because no image data is sacrificed to create the compressed version.
Lossy compression methods such as JPEG or indexed color formats actually remove image data in order to
decrease file size; this can affect a digital watermark’s durability.

So for being "interwoven", it can be actually be made unreliable or completely destroyed. They use an almost invisible "change" to the image in order to plaster their watermark shit all over it, almost like a dog with invisible pee that can be smelled by other dogs...

I do get the importance of watermarking as I read through some forums where musicians said they lose out on so much money. I often thought about my favourite composers making extremely high-quality music, press the CD, comission cover art and maybe two licensees actually pay for it? But man, these people are obsessive and I think you can overdo it. Normally, people want to rip out any kind of DRM out of their pirated files, but if it becomes too obnoxious, people won't enjoy it. Did it help the music industry about CDs? Did it help that BluRays are a pain in the ass to rip even for your legal rights (personal backup copy)? iTunes says you are not allowed to remove DRM, but I did, because I have the legal right of a "platform-independent backup copy", and making it "platform independet" means to rip all "Apple, Inc." out of it. You can even purchase DRM remover legally, but they still added a higher bar, smirking: "Go ahead then", basically denying your legal rights...

I mean, imagine you record a track from the radio and it has this Digimarc jizz all over it as it apparently survives everything like a cockroach. Sorry, f*ck off. You got your rights and money from the broadcast station, so get your s*hit off it, thanks. Would you want it, pirate or legit user or not? I sure would not! Thankfully, my collection is mostly made of CD rips and LP recordings, but the newly added digital releases might have Digimarc all over it. They were apparently already getting active in 2018, but the acceptance of Digimarc wasn't great back then ... I heard they also have ASCAP and BMI and all the right performers on board, so the newly released music might be "infected". In itself not a problem as you don't broadcast, but still, to close up my post:

No, it is NOT lossless anymore!

Yeah, so, in the future, when someone on this forum asks me for help with some of these wavs.

I'll prioritize other sites, and when I can't, I'll choose SourceAudio.

But no matter what, it's still a WAV, so it's better than an MP3.

If they say we can't hear the watermarks, then it must have been rigorously tested.

Unless, of course, you have a very strong mental fetish, or a mental suggestion.

But what is certain is that more and more libraries are choosing to work with SourceAudio.

[soundtech39]

So my advice is that WAV on CD is the best.

What do you think?

I will always rip from CD.  Unless I can't find that CD or it's unreasonably priced.

[kpmhill]

iTunes says you are not allowed to remove DRM

There's no DRM on purchased tracks from Apple Music.