Music Meets Steganography

austad writes "Wired is running a story about how Aphex Twin has encoded a face into one of his songs. The face is visible when viewing the sound through a spectrograph. This is probably something I wouldn't want to see when coding in a dark room at 3AM. Sorry boys and girls, you have to buy the CD if you want to see it, encoding of the song into a lossy format destroys the image."

Comments

no it doesn't

[Firlefanz]

encoding of the song into a lossy format destroys the image."

I have an mp3 encoded @192kps, using the Nullsoft tiny fullscreen plugin displays the image just fine (its at the last few seconds of the 2nd track of the Windowlicker EP.

Re:no it doesn't

[CaseyB]

I can confirm this. I tried it out when I first heard the story a month or two ago. It's not a checksum or some other digitally embedded data block, it's simply the spectrograph output. It doesn't change due to compression, it will just gradually degrade. MP3 quality is way more than adaquate.

A few of his other songs do similar things, with spirals and other designs appearing in the spectrograph. But the windowlicker track contains a digitized image. Very cool.

aphex twin

[tps12]

Man, if ever there was a musician who deserves the geek limelight (geeklight?), it is AFX. Not only does he hack his instruments and work primarily (solely?) with homebrewed samples, but he has a fucking tank.

Also, his music is amazing.

Re:You should spell-check your title lines

[PCM2]

It's Stenography.

No, it isn't.

Re:A way to boost sales...

[Graspee_Leemoor]

Yeah, but if you rip to .flac then you don't lose anything, so you could pirate and still see the face. With people using 700MB for a divx movie, sometimes even 1.4Gig, using 250MB for a .flac album is not really very much space...

graspee

Re:A way to boost sales...

[CaseyB]

Interesting point, but irrelevant, because that part of the slashdot story is completely false. (Surprise, surprise, I know) The image is almost entirely unaffected by compression.

I wrote a tool to hide text in spectrograms

[yerricde]

I once wrote a program to hide printed text in a spectrogram. The first thing I encoded (after test messages such as Hello World) was efdtt from David Touretzky's Gallery of CSS Descramblers.

the program

efdtt on top of music from Tet*is Advance

Use 64 kbps

[yerricde]

What ripping quality would preserve the face?

Steps used in Cool Edit Pro with Fraunhofer plug-in:

1. Rip CD
2. Trim to face only (the face looks strange in a linear spectrograph such as the one in Cool Edit)
3. Convert to Mono
4. Save as Fraunhofer MP3 at 64 kbps
5. Open MP3 in Winamp
6. Turn on Nullsoft Tiny Visualizer and play the MP3. The face is preserved, but unfortunately, Winamp's spectral display is linear too.

Anybody have a good link to a spectrograph program that uses a logarithmic frequency axis?

--
DeCSS hidden in a song's spectrogram

Re:Use 64 kbps

[JesseL]

This probably won't do you a lot of good since you seem to be running windows, but, Extace Waveform Display (came with RedHat 7.2) does have an option for using a logarithmic frequency axis.

It's not steganography

[rhizome]

Steganography is encoding something in another medium so as not to alter the carrier medium, like a watermark. What Aphex Twin did was to use a piece of software that converts graphics to sound (x, y, z = time, frequency, and intensity/volume) via an Inverse Fast-Fourier Transform. There is no encoding involved, the picture *is* the sound that you hear. Big difference.