Re: You have a problem here. All the feds have to do is go before the judge, start with an archived copy of the original disk which can be proven to be bitwise identical, apply the correct decryption process, and when out pops all the data they claim was there you'll have to explain how they got the original encrypted bytes to decrypt directly into the alleged criminal data.
I can make you a one-time pad that can decrypt any X of length $L_x$ into any Y also of length $L_x$. Just fucking XOR the values of X and Y and call that Z = (X) xor (Y) . Now, Z is the XOR-decryption-encryption key that will transform X into Y. That can be done for every fucking Y possible in the world. Of course, this is charlatanry. You're not going to use a one-time pad of length 1-terabyte, or are you?
But it's trivial. Say that my code is "12345". Say that I want to decode that into "fuxor". Set Z = "12345" xor "fuxor", and say that Z is the encryption key. Now, (Z) xor ("12345") = "fuxor". Tada. La Voila.
But what if instead I wanted to prove that "12345" really encoded "itsCP" ?? Well, claim that Z_2 = "12345" xor "itsCP". Now, (Z_2) xor ("12345") = "itsCP". Ohmifuckinggod, the code "12345" decodes into the obvious claim that "itsCP"!!! There can be no other answer!!! Hang'em!
And that can be done for the noisy chaff that is encrypted data too. It's like the Queen in Alice Wonderland explaining what words mean: "I can make them mean anything I want them to mean!"
in other words, tl;dr: I can make a one-time pad to prove it's anything...