I was thinking about a very simple steganographic technique to simulate the noise floor for lossless image formats. It assumes that the input bitstream is already encrypted and has uniform distribution. The algorithm is as follows: Divide the image into 2x2 pixel blocks. Consider the MSB of each pixel in a block. If it's not ((1,0),(0,1)) or ((0,1),(1,0)), then skip to the next block. Take a bit from the input; if it is 0, replace the block's pixels' MSBs with ((1,0),(0,1)); otherwise, replace them with ((0,1),(1,0)). Repeat the process until you run out of input bits. The main disadvantage of this technique is, of course, that the vessel data needs to be 32 times larger than the signal. But at least it should be very hard to detect.
>>2
I don't get it.