Tidal and Watermarking (and MQA) ?

[PeterSt]

 

19 hours ago, matthias said:

Is there watermarking on streamed ECM music from Tidal and Qobuz?

Thanks

 

Matt

 

8 hours ago, matthias said:

There is a comment from Audiostream about audible watermarking on ECM streaming:

 

https://www.audiostream.com/comment/517715#comment-517715

 

Matt

 

Quote of the text referred to :

 

Unfortunately a lot of the ECM titles available on Tidal still suffer from the effects of an attempt to brand the material. Called watermarking and allegedly performed by Universal to some/all of their material.
If you havent heard it, give Nik Bärtch’s Ronin - Stoa a listen. On the first song, Modul 36, you can hear this clearly when the piano emerges from the pitch black background and then fades away. Someone described the effect as listening through a ceiling fan?
Too bad, on such excellent recordings!

 

8 hours ago, matthias said:

Maybe Tidal allows watermarking on redbook content to promote the MQA version of the same music.

 

I thought to give this some emphasis in a dedicated thread, because if true ... NOT good.

 

Initially the comment is about the leftmost album you see below, first track :

 

 

The other two we can compare to see whether the phenomenon is present in the MQA version of it as well.

 

Watch out : I can already see that the first pair for comparison is *not* from the same master (fed to MQA's processing).

The second, rightmost pair *is* from the same master, but obviously only for the 4 tracks of the first one (01, 02, 03, 04) which are contained in the second (MQA) one (06, 07, 08, 09. A bit confusing, but if you see it, you will understand.

 

I will try to compare a few things myself, but (much) later today.

Peter

 

 

 

 

 

 

[matthias]

Peter,

thanks for joining the posts in a new thread.

I also hope that this all is not true.......

 

Matt

[PeterSt]

Hi Matt,

 

10 hours ago, matthias said:

I also hope that this all is not true.......

 

Do you have more examples from forums etc. that you know of ? I think that the guy who came up with this one was listening in his car or had the hairdryer on or whatever, because for someone who plays synths and when playing the grand piano makes flageolets with it including all sorts of sounds, well, I can't make up what I'm hearing in Modul 36 except for a most normal "manipulative" sound. Let's keep in mind that my interest is in electronic music so it is relatively easy for me to recognize.

But merely : anyone who judges this kind of music as watermarked because he hears a strange sound ? well, what to say. But maybe when compared with vinyl or something but then still (digital digs out quite a bit more from a recording than LP ever can).

 

Thanks,

Peter

 

[matthias]

Hi Peter,

 

to get a bigger picture I think it is mandatory to  compare the same piece of music

 

1.) on Tidal vs. Qobuz

and/or

2.) on redbook vs. MQA vs. high-res (Qobuz) vs. CD vs. LP (if available)

 

When I have more examples I will post them here.

 

Thank you

 

Matt

[PeterSt]

10 minutes ago, matthias said:

When I have more examples I will post them here.

 

That would be great, Matt.

Anyway, I am still listening to this superb music (now at the rightmost one in that row) and the MQA-44 sounds the best here (that one is live). really great stuff and I should thank you for it.

So Thank You.

 

Edit : I forgot ... and it is manipulation of the the piano wings continuously. And other "sounds" all over. My style !

[Don Hills]

Homework for you, especially the first link:

 

http://www.mattmontag.com/music/universals-audible-watermark

 

http://www.mattmontag.com/music/weird-spotify-compression-artifacts

 

http://research.microsoft.com/pubs/102066/kirovskimalvartspapr03.pdf

 

http://www.hydrogenaudio.org/forums/index.php?showtopic=89818

 

http://www.techdirt.com/articles/20110809/04114515451/umg-watermarks-audiophile-files-pisses-off-paying-customers.shtml

 

https://www.eff.org/press/mentions/2008/1/11-0

 

 

[PeterSt]

2 minutes ago, Don Hills said:

Homework for you, especially the first link:

 

Oh wow. 

And this is not to be called DRM of some sort ?

 

Thank you, Don.

[Miska]

Usually they'd want to watermark subscriber to the stream data, so any illegal copies can be traced back. I'm pretty sure MQA can do the same too, along with their other metadata.

 

One could dump same Tidal track from two separate accounts and check if they match or not. Easy to confirm.

 

[left channel]

This is bigger than Tidal. From first link provided by @Don Hills:

 

"It turns out the artifacts I was hearing are not due to compression, but a result of audio watermarks that Universal Music Group embeds in digitally distributed tracks. This watermark is embedded in UMG tracks on Rdio, Spotify, iTunes, Tidal, and others. The watermark can also be heard in Universal tracks broadcast over FM radio. Universal Music recordings make up about 25% of most online catalogs..."

https://www.mattmontag.com/music/universals-audible-watermark z

 

Even FM radio??? Sigh.

[FredericV]

2 hours ago, Miska said:

Usually they'd want to watermark subscriber to the stream data, so any illegal copies can be traced back. I'm pretty sure MQA can do the same too, along with their other metadata.

 

One could dump same Tidal track from two separate accounts and check if they match or not. Easy to confirm.

 

The flac files on Tidal have a very simple token protection system that limits access to a file on Tidal for a certain time window for a given token. When accessing those files via chrome, they are encrypted. When accessing them via open source players such as LMS, the flacs are not encrypted. So for every flac, they have at least 2 versions on their CDN, or the webserver is encrypting on the fly. Note that this is not HTTPS/SSL. The HTTP response header of the encrypted version is still plaintext, while the http body is encrypted.

I figured out the token system when I was playing with a track in repeat, it would re-use the same token according to the http headers in the debug output of a player. It certainly does not use one time tokens.

I don't believe Tidal is encoding flacs per user, with per user watermarking. That would imply processing the flac per user, which would cost a lot of resources on their CDN.

[Miska]

2 hours ago, FredericV said:

I don't believe Tidal is encoding flacs per user, with per user watermarking. That would imply processing the flac per user, which would cost a lot of resources on their CDN.

 

I don't know, FLAC encoding is not too bad. Obviously they are not doing it at the client end. But it could be checked.

 

iTunes at least seemed to do watermarking at the client end when downloading purchased AAC's. I haven't used it in years so I don't know how they deal with it now.

 

I don't see a point in doing global watermarking where the watermark is the same always...

 

[mansr]

21 minutes ago, Miska said:

I don't see a point in doing global watermarking where the watermark is the same always...

Music industry paranoia knows no bounds.

[left channel]

Why does this conversation keep coming back to Tidal? Based on what I posted above, watermarking is being done by UMG labels across all streaming/broadcast platforms, including FM radio, and this is measurable as well as audible. And yes, OK, MQA is a thing. But I can't find any proof that Tidal is doing additional watermarking of any kind on their own. Am I missing something?

[Miska]

2 hours ago, left channel said:

Why does this conversation keep coming back to Tidal? Based on what I posted above, watermarking is being done by UMG labels across all streaming/broadcast platforms, including FM radio, and this is measurable as well as audible. And yes, OK, MQA is a thing. But I can't find any proof that Tidal is doing additional watermarking of any kind on their own. Am I missing something?

 

I wouldn't be surprised if a big label demands to watermark buyer/subscriber information as a condition for allowing their content on the service. Usually they'd already have some preferred solution too. Like "if you can use watermarking provider X, we can make our catalog available to you". I can very well imagine that UMG could have their own CDN and Tidal would be just acting as a proxy between their backend and customer, where all watermarking would be done by the label's CDN and Tidal wouldn't even know about it. Supporting such private CDN would require backend customization from Tidal, but I can very well imagine such being not a problem for such a big label.

 

I am not claiming Tidal is doing this, but just speculating what could be happening. Because that way I see more point in such exercise than just providing bunch of files for Tidal that have single watermark (what use would it be?).

 

As a method to check this, one could download the same FLAC from Tidal using two different user accounts and compare if the SHA-256's match...

 

[PeterSt]

58 minutes ago, Miska said:

As a method to check this, one could download the same FLAC from Tidal using two different user accounts and compare if the SHA-256's match...

 

They match. But I am not sure to what degree this all influences in the context of what you literally asked and what I did, Miska :

 

10 hours ago, FredericV said:

The flac files on Tidal have a very simple token protection system that limits access to a file on Tidal for a certain time window for a given token. When accessing those files via chrome, they are encrypted. When accessing them via open source players such as LMS, the flacs are not encrypted. So for every flac, they have at least 2 versions on their CDN, or the webserver is encrypting on the fly. Note that this is not HTTPS/SSL. The HTTP response header of the encrypted version is still plaintext, while the http body is encrypted.

I figured out the token system when I was playing with a track in repeat, it would re-use the same token according to the http headers in the debug output of a player. It certainly does not use one time tokens.

I don't believe Tidal is encoding flacs per user, with per user watermarking. That would imply processing the flac per user, which would cost a lot of resources on their CDN.

 

So let's be careful with drawing the right conclusions.

[FredericV]

I monitor & log all the network traffic, here are all the Tidal hosts my players tried to get flac files from:
 

http://00.audio-pop.tidal.com
http://02.audio-pop.tidal.com
http://03.audio-pop.tidal.com
http://04.audio-pop.tidal.com
http://05.audio-pop.tidal.com
http://06.audio-pop.tidal.com
http://07.audio-pop.tidal.com
http://09.audio-pop.tidal.com
http://0b.audio-pop.tidal.com
http://0c.audio-pop.tidal.com
http://0d.audio-pop.tidal.com
http://0e.audio-pop.tidal.com
http://10.audio-pop.tidal.com
http://11.audio-pop.tidal.com
http://12.audio-pop.tidal.com
http://13.audio-pop.tidal.com
http://14.audio-pop.tidal.com
http://15.audio-pop.tidal.com
http://16.audio-pop.tidal.com
http://17.audio-pop.tidal.com
http://18.audio-pop.tidal.com
http://19.audio-pop.tidal.com
http://1a.audio-pop.tidal.com
http://1b.audio-pop.tidal.com
http://1c.audio-pop.tidal.com
http://1d.audio-pop.tidal.com
http://1e.audio-pop.tidal.com
http://1f.audio-pop.tidal.com
http://20.audio-pop.tidal.com
http://21.audio-pop.tidal.com
http://22.audio-pop.tidal.com
http://23.audio-pop.tidal.com
http://24.audio-pop.tidal.com
http://25.audio-pop.tidal.com
http://26.audio-pop.tidal.com
http://27.audio-pop.tidal.com
http://28.audio-pop.tidal.com
http://29.audio-pop.tidal.com
http://2a.audio-pop.tidal.com
http://2b.audio-pop.tidal.com
http://2c.audio-pop.tidal.com


...

http://f1.audio-pop.tidal.com
http://f2.audio-pop.tidal.com
http://f3.audio-pop.tidal.com
http://f4.audio-pop.tidal.com
http://f5.audio-pop.tidal.com
http://f6.audio-pop.tidal.com
http://f8.audio-pop.tidal.com
http://f9.audio-pop.tidal.com
http://fa.audio-pop.tidal.com
http://fb.audio-pop.tidal.com
http://fc.audio-pop.tidal.com
http://fd.audio-pop.tidal.com
http://fe.audio-pop.tidal.com
http://ff.audio-pop.tidal.com

No proof (yet) of specific CDN's for certain labels.

The URL is always like this:

GET http://d1.audio-pop.tidal.com/4f45019150aea3eb89782667db0125b2_26.flac?__token__=exp=1502736598~hmac=string

where string is  a 64 char hex string

[Miska]

5 hours ago, FredericV said:

No proof (yet) of specific CDN's for certain labels.

 

Yeah, but that doesn't prove anything. You wouldn't know if Tidal is just acting as a proxy.

 

Typically those nodes are just web accelerators, the whole point of CDN. I use such myself too. But the actual hosting behind it is likely somewhere completely elsewhere. So those nodes usually don't even have disk space, just small SSD to boot up and everything else runs straight out of RAM. But they could handle on-the-fly watermarking. I'm pretty sure I could implement pretty fast parallelized FLAC watermarking myself using GPUs. Doing like 1000+ streams in parallel.

 

And you could multiplex the requests to several distinct second-tier CDNs based on hashes included in the URLs. In fact, one would probably make it tree-like structure with few layers.

 

5 hours ago, FredericV said:

The URL is always like this

 

I know, because I've been playing Tidal using HQPlayer Embedded...

 

[left channel]

We should be focusing on the source of this problem: UMG. Apparently they've been implementing this years and yet have received no pushback of consequence. General suspicion of Tidal and the other streaming services is a distraction. Of course they will succumb to pressure from labels before they listen to us. The executives of UMG have nothing but contempt for us. Here is a patent for inaudible watermarking that they have ignored: https://www.google.com/patents/US7266697?dq=7,266,697 And if you're concerned about DRM and privacy? Again, I say we should direct our outrage at the labels first. Congressional hearings on the potential misuse of watermarking and MQA would be a nice start.

[Miska]

Yeah, I don't think any streaming provider wants to do watermarking, but they could be pressured to do so by labels. Or they may not even specifically know about it if they are just proxies... Like "we just relay data between label X and customer A".

 

MQA is kind of example on watermarking, whether it is audible or not in unfolded or non-unfolded variants is up to everybody to decide.

 

Reason why many of those watermarkings are audible is that they are designed to survive MP3 encoding to something like 128 kbps. Pretty much meaning that it needs to be audible in order avoid MP3 encoder from discarding it...

 

[left channel]

I've seen nothing to indicate the streaming providers are doing watermarking. The music apparently arrives that way from UMG, and maybe from others. And since inaudibility is an important goal of any watermarking, the fact that it is so badly done by UMG is particularly galling.

[PeterSt]

Hmm ... It could be a good advice to not find out how this sounds because I can imagine that once you know it you hear it forever.

I am still virgin regarding this. 

[PeterSt]

Does anyone have experience with "a best" detector regarding this ? I didn't Google on it yet, but I have the idea that @left channelknows a thing or 2 about this. This is also about this, what I announced already :

 

9 hours ago, Miska said:

MQA is kind of example on watermarking, whether it is audible or not in unfolded or non-unfolded variants is up to everybody to decide.

 

but put differently from what Miska tells there ... *if* a label does the watermarking then it could very well be so that for the MQA versions it's done in the MQA meta data band and then we a. won't see it easily and b. won't hear it.

I would imagine that such a detector won't find the watermarking in MQA either, but for me it would be sufficient that it won't detect it where it otherwise would detect it in the normal music band(s) and so we wont hear it in MQA, never mind it is in there.

[left channel]

@PeterSt I only know what I've read. And what I've heard here http://www.mattmontag.com/music/universals-audible-watermark

I think we need to separate watermarking we *know* is happening from suspicions of MQA and streaming services in general.

 

 

[PeterSt]

@left channel, OK thanks.

By now I did Google and I don't have hope to find a "detector" if you first don't know which software created it. If we know that, it is a matter of buying or perhaps trialling that software, which always comes along with a "reading back" feature (see if the audio file is yours, with your watermark in it).

 

Do we know what UMG puts in there ? Maybe @Don Hillssaw it passing by somewhere ?

[left channel]

@PeterSt listen to the samples at this link provided by @Don Hills: http://www.mattmontag.com/music/universals-audible-watermark