Music Steganography

A lot of the more recent music on this site (2007 and later) contains bits and pieces of hidden information in one form or another. This is a form of steganography; in the case of the music it's more of a complex "signature" rather than a form of covert communication.

The idea came to me after a project where I was attempting to convert audio from old PXL-2000 tapes, which are audio cassettes containing low-resolution video on the left channel and audio on the right channel.

Capturing the audio (actually containing the camera's specially-formatted video as well) using a high sample rate with the computer's sound card yielded relatively poor results as most of the high frequency analog signal had degraded into almost nothing over the ~20 years the tapes had been in storage. It did, however, spark some new ideas as I read the captured audio in as a raw bitmap (PCM audio and grayscale bitmap images are essentially the same thing).

Noticing that images and audio can be converted to each other, I started converting images to audio and mixing the audio in at low volumes in audio files of, well, actual audio. The result was a faint imprint of the image in the sound file, which could later be extracted and viewed - albeit a very noisy (the "noise" being the other audio in the mix) copy of the original image. It's very much like a form of watermark.

After doing some searching on the Internet for other forms of what I was experimenting with, I learned that Aphex Twin had been doing similar things.

After obtaining my amateur radio license in 2007, I also started experimenting with PSK 31 and slow-scan television. PSK31 can be set up to use different frequencies and I tend to tune it to match the key of the song. It blends in well. Slow-scan television doesn't blend in so well so when it's hidden in there, it's usually very quiet.

Another trick to apply hidden "signature" data in songs is to record a voice message at a low sample rate (usually 8 kHz), speed up the recording by changing the sample rate to match the song, making the voice message very quiet, then mixing it in at the very end of the song after the fadeout. This is quite common in the music section.

Finally, after doing quite a bit of searching, I found some software that encodes images images in a song in such a way that they can be viewed using a sonogram. This is basically what Aphex Twin uses and produces a relatively easy to view result. It's certainly easier than converting the audio to a raw bitmap and opening it in an image editor - you would have to guess the dimensions of the image in order to view it - but using the sonogram method requires audio editing software that generates sonograms. Of course, you can always generate a sonogram using sox if you don't mind using the command line interface.

So, enjoy hunting around for the different Easter eggs in the music; I highly recommend playing around with the aforementioned software.