Is AES statistically noise? Reliably so?


Suppose we want to store files in a way that has plausible deniability. Ie. we can deny the existance of those files.

The approach is this: We allocate (or just use) a chuck of disk an initialize it with static noise. Then we write our file in there encrypted. We take care to handle the beginning of the ciphertext properly, so we are interested if an attacker can find out the place to look for this file, can they tell it is not in fact pure noise, but ciphertext?

My current implementation also includes a randomized hopping pattern around the chunk derived from the same encryption key. The purpose of this is not only to obfuscate the ciphertext more, but also to distribute the data evenly (all bytes are written 3-9 times) so that subsequent storage operations do not overwrite previous data so easily. I was wondering if one uses a more deterministic and public storage pattern, could a ciphertext be distinguished from noise if correct sequence is discovered?