09.05.08
Audio Steganography
I was inspired by Ohmpie’s post, “Encoding an image to sound” about sound in gray scale images. One extension of his work is steganography. Steganography aims to send a message in an image without others’ knowledge that an encrypted message is attached. I started with the following image, upon which I could add data.
The following image has hidden audio data encoded in the colors used in the image.
The audio encoded in the image was extracted from this image. You can listen to it.
Audio after decryption
A closer look at the image shows some of the dithering caused by the extra data.
How does it work?
For this project, I worked with X-Pixmap (XPM) files, which are trivial to manipulate in Gimp and Perl. I sorted the colors using a 3-D Hilbert curve. It had to be 3-D curve to account for red, green and blue. The Hilbert curve guarantees that not one color will differ from the next significantly. I sorted the entire spectrum into one dimension. I only used one out of every 16 colors in the palette. This leaves me with 4 bits per pixel. For this image size, this allows me to embed 34kB of data of any type. For embedding the hidden data, I sorted the colors based on the scheme described above, then dithered the color slightly to add the 4 bits. To decode, I sorted the color palette, computed the length along the Hilbert curve, modulo 16, then took the remainder, which I concatenated to re-create the data.
The code for encoding and decoding is here.
Finally, this image contains this entire blog post.
