IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Fhg's "inverse Decoder" - Or Encoder Detector
Ivan Dimkovic
post Sep 8 2002, 17:48
Post #1


Nero MPEG4 developer


Group: Developer
Posts: 1466
Joined: 22-September 01
Member No.: 8



I've just read interesting AES article (#5256) regarding FhG's "Inverse Decoder" tool..

It seems that this tool is able to:

- Analyse the encoding window size (1152 / 1024 / ... samples)
- Analyse the short block usage
- Analyse the bit reservoir usage
- Analyse the M/S and I/S usage
- Analyse the quantization step size
- Analyse the TNS usage

With this information, they could detected what algorithm was used, and with database of known properties - what encoder and version, as well as encoding settings were used only by analysing of decoded PCM file!!

This is very interesting and straightforward algorithm for transcoding without losses, and they claim that they can even reconstruct MP3/AAC bitstream out of decoded WAV without degradation of sound quality!

But, there are some other usages, like this:

QUOTE
The increasing popularity of low bitrate audio coding schemes comes along with a significant number of undesirable activities, such as the illegal distribution of copyrighted audio over the Internet and the availability of encoders illicitly using patented algorithms. In this context, the inverse decoder can be an interesting tool both for forensic purposes as well as for investigations about violations of IPRs.



So, the tool could be used for detecting unlicensed MP3 and AAC encoders, or illegally obtained MP3 encoders smile.gif
Go to the top of the page
+Quote Post
AgentMil
post Sep 8 2002, 17:52
Post #2





Group: Members (Donating)
Posts: 584
Joined: 19-December 01
From: Australia
Member No.: 688



Ivan, this is very very interesting! So smart these people are! biggrin.gif

Laters
AgentMil


--------------------
-=MusePack... Living Audio Compression=-

Honda - The Power of Dreams
Go to the top of the page
+Quote Post
JohnV
post Sep 8 2002, 17:54
Post #3





Group: Developer
Posts: 2797
Joined: 22-September 01
Member No.: 6



I'm wondering if somebody could implement a transcoder detector, which would detect from AAC,Vorbis or MPC -file if the source was something like MP3 or WMA.. tongue.gif


--------------------
Juha Laaksonheimo
Go to the top of the page
+Quote Post
Ivan Dimkovic
post Sep 8 2002, 17:55
Post #4


Nero MPEG4 developer


Group: Developer
Posts: 1466
Joined: 22-September 01
Member No.: 8



I was really impressed by reading the article (by Jürgen Herre - one of co-developers of MP3 and AAC standards) - it is really very interesting and straightforward.

Too bad it could be used for "corporate business" (filing lawsuits all around)
Go to the top of the page
+Quote Post
Frank Klemm
post Sep 8 2002, 19:04
Post #5


MPC Developer


Group: Developer
Posts: 543
Joined: 15-December 01
From: Germany
Member No.: 659



QUOTE (Ivan Dimkovic @ Sep 8 2002 - 06:55 PM)
I was really impressed by reading the article (by Jürgen Herre - one of co-developers of MP3 and AAC standards)  - it is really very interesting and straightforward.

Too bad it could be used for "corporate business" (filing lawsuits all around)

Take a microphone and convert the article into an Ogg Vorbis file ;-)


--------------------
-- Frank Klemm
Go to the top of the page
+Quote Post
wkw
post Sep 8 2002, 19:07
Post #6





Group: Members
Posts: 85
Joined: 7-June 02
Member No.: 2241



This inverse decoder isn't hard to implement.. In fact in the development of my encoder, I generated a lot of debuging data... For every 1 MB of pcm audio compressed.. the encoder generates 30 - 40 MB of debugging data.. All that is needed is to move these codes into the decoder... and hey.. you have an inverse decoder.. In fact.. using the Cool Edit tools, you find out when exactly at what audio samples did the encoder switch to start / short / stop block! You even know exactly when and where TNS is enabled and best of all, you can zoom in zoom out using the Cool Edit graphical zoom capabilities..

However, I doubted much can be done about the psychoacoustic model settings.. There is no way to extract psychoacoustic model settings from the decoder...

wkw
Go to the top of the page
+Quote Post
Frank Klemm
post Sep 8 2002, 19:11
Post #7


MPC Developer


Group: Developer
Posts: 543
Joined: 15-December 01
From: Germany
Member No.: 659



QUOTE (Ivan Dimkovic @ Sep 8 2002 - 06:48 PM)
I've just read interesting AES article (#5256) regarding FhG's "Inverse Decoder" tool..

It seems that this tool is able to:

- Analyse the encoding window size (1152 / 1024 / ... samples)
- Analyse the short block usage
- Analyse the bit reservoir usage
- Analyse the M/S and I/S usage
- Analyse the quantization step size
- Analyse the TNS usage

With this information, they could detected what algorithm was used, and with database of known properties - what encoder and version, as well as encoding settings were used  only by analysing of decoded PCM file!!

This is very interesting and straightforward algorithm for transcoding without losses,  and they claim that they can even reconstruct MP3/AAC bitstream out of decoded WAV without degradation of sound quality!

But, there are some other usages, like this:

QUOTE

The increasing popularity of low bitrate audio coding schemes comes along with a significant number of undesirable activities, such as the illegal distribution of copyrighted audio over the Internet and the availability of encoders illicitly using patented algorithms. In this context, the inverse decoder can be an interesting tool both for forensic purposes as well as for investigations about violations of IPRs.



So, the tool could be used for detecting unlicensed MP3 and AAC encoders, or illegally obtained MP3 encoders smile.gif

IMHO a tool is possible which determines the used encoder from the decoded PCM data
(which is much more difficukt than determine this from the data stream).

But to determine the used encoder is much much much more simple than a inverse decoder.
To my mind an inverse decoder is not possible for current computers, and actually not necessary.
If you need an inverse decoder you have actually a coordination problem which can also
be solved by solving this corrdination problem.

BTW there are also aproaches for inverse decoders using watermarking in the decoder.
Especially with current pop music this works fine, because you don't need to 1...2 LSBs of the
decoded PCM signal.


--------------------
-- Frank Klemm
Go to the top of the page
+Quote Post
layer3maniac
post Sep 8 2002, 19:17
Post #8





Group: Banned
Posts: 529
Joined: 29-September 01
Member No.: 37



To HECK with the OTHER uses. THIS is what caught MY eye!
blink.gif
QUOTE (Ivan Dimkovic @ Sep 8 2002 - 08:48 AM)
This is very interesting and straightforward algorithm for transcoding without losses,  and they claim that they can even reconstruct MP3/AAC bitstream out of decoded WAV without degradation of sound quality!
Go to the top of the page
+Quote Post
wkw
post Sep 8 2002, 19:25
Post #9





Group: Members
Posts: 85
Joined: 7-June 02
Member No.: 2241



I don't how such a tool can "remodelled" the psychoacoustic masking threshold?
You can only extract the G_Gain & scalefactors... But to reconstruct the psychoacoustic model settings from these informations is near impossible.
Further more, these algorithm may not even be using the standard psychoacoutic model II ??? It may be some proprietary psychoacoustic model..

wkw
Go to the top of the page
+Quote Post
DSPguru
post Sep 8 2002, 19:31
Post #10


BeSweet Developer


Group: Developer
Posts: 235
Joined: 14-December 01
From: sitting on top of a supercompact cardinal
Member No.: 643



QUOTE (Frank Klemm @ Sep 8 2002 - 09:11 PM)
Especially with current pop music this works fine, because you don't need to 1...2 LSBs of the decoded PCM signal.

hihi laugh.gif


--------------------
DSPguru's Webpage - home of BeSweet & The OggMachine.
BeSplit - 7kb of A loseless AC3/AAC/MP3/MP2/WAV Cutter :)
Go to the top of the page
+Quote Post
Ivan Dimkovic
post Sep 8 2002, 20:01
Post #11


Nero MPEG4 developer


Group: Developer
Posts: 1466
Joined: 22-September 01
Member No.: 8



No, of course it is not possible to guess psych. model settings,

but with all settings tool could guess, FhG could easily build database of all evailable encoders and their settings for each bit rate - and this way they could detect what encoder was used and with what default settings.
Go to the top of the page
+Quote Post
Frank Klemm
post Sep 8 2002, 20:04
Post #12


MPC Developer


Group: Developer
Posts: 543
Joined: 15-December 01
From: Germany
Member No.: 659



QUOTE (Ivan Dimkovic @ Sep 8 2002 - 09:01 PM)
No, of course it is not possible to guess psych. model settings,

but with all settings tool could guess, FhG could easily build database of all evailable encoders and their settings for each bit rate - and this way they could detect what encoder was used and with what default settings.

Now I know why lame has soooo much options.


--------------------
-- Frank Klemm
Go to the top of the page
+Quote Post
Ivan Dimkovic
post Sep 8 2002, 20:11
Post #13


Nero MPEG4 developer


Group: Developer
Posts: 1466
Joined: 22-September 01
Member No.: 8



Hehe, right :-)

And, as far as I understand - some companies are distributing LAME legally (by paying $15k/yr) and therefore FhG could never tell was the distributed LAME encoding "legal" or not smile.gif
Go to the top of the page
+Quote Post
rjamorim
post Sep 8 2002, 21:19
Post #14


Rarewares admin


Group: Members
Posts: 7515
Joined: 30-September 01
From: Brazil
Member No.: 81



QUOTE (Ivan Dimkovic @ Sep 8 2002 - 04:11 PM)
some companies are distributing LAME legally (by paying $15k/yr)

MusicEx/JRiver is distributing LAME legally with MediaJukebox.


--------------------
Get up-to-date binaries of Lame, AAC, Vorbis and much more at RareWares:
http://www.rarewares.org
Go to the top of the page
+Quote Post
Frank Klemm
post Sep 8 2002, 22:28
Post #15


MPC Developer


Group: Developer
Posts: 543
Joined: 15-December 01
From: Germany
Member No.: 659



QUOTE (Ivan Dimkovic @ Sep 8 2002 - 09:11 PM)
Hehe, right :-)

And, as far as I understand - some companies are distributing LAME legally (by paying $15k/yr)  and therefore FhG could never tell was the distributed LAME encoding "legal" or not smile.gif

And what about Xing encoded files re-encoded with FhG?

(I currently sorting nearly 20 CD-Rs with good music, but 60% is encoded
with Xing, Simple Stereo, 128 kbps, 25% with ISO-Encoder (wrong CRCs!),
Simple Stereo, 128 kbps, the rest with FhG or >=160 kbps).

I would say the main problem with MP3 is that most of the files are encoded with crappy
encoders.


--------------------
-- Frank Klemm
Go to the top of the page
+Quote Post
Gabriel
post Aug 26 2003, 12:00
Post #16


LAME developer


Group: Developer
Posts: 2950
Joined: 1-October 01
From: Nanterre, France
Member No.: 138



link:
http://www.sci.utah.edu/~sascha/pubs/Inver...erseDecoder.pdf
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 31st July 2014 - 08:06