This work have an interesting history. Someone asked me to write a cryptosystem. I'd analysed some exists ones and disclouse, that all of them can be cracked. Ok, nobody know how to do this right now, but...
But over the couple of years somebody will arrange the system of equations and solve it on the computer.
Now all the algorithms, used in serious cryptosystems are trans-computable. I'd decided to create trans-undrstandable one.
As long as compression algorithms already used in cryptosystems, and as long as they are very complicate, I use they as a based bricks for my cryptosystem.
But how we can use the compression algorithms in cryptosystem?
We can commutate two or more of them or (and) change some parameters of the compression algorithms dynamically under control of the special pseudo-random generator of the cryptosystem.
In my real cryptosystem I commutate LZSS+Arithmetical coder with LZSS+Huffman coder. It seems to be impossible to keep in mind (and analyse) all this monster together. So, this algorithm seems to be either trans-understandabe or very-very close to.
Of cause, I can offer a number of another solutions using, for example, PPM or some types of image compression. Moreover, now I can rewrite my cryptosystem and make it twice more easier (Now I know a better way for synchronizing Huffman with Ari) and I know how to make twice more compliate scheme.
The only one question - are we really need a
Or trans-computable ones is enough for us?