new idea for random number generation

[dreamingforward at gmail.com (\0xDynamite)]

>
>  >>Hmm, as far as I'm aware, you can't do cryptanalysis on 2 characters.
>>>Marxos
>>> A factoid which is irrelevant in this context.
>>> If you only wanted to encrypt two characters, once, you could use ROT-13
>>> and
>>> nobody would be the wiser.
>
>>ot a factoid.  I'm making the claim that you can't decrypt it.  I
> don't want to do the proof, because I can deduce it, with Pretty Good
> Certainty (TM).  If you have a key approx 1/2 the size of the text,
> AND you do NOT know the length of the key, you have essentially two
> characters to do cryptanalysis.  Now the key may be a small enough
> dictionary to do more analysis, and since you know that each set of
> the same modulus is using the same key character, but without knowing
> the length of the key, it's going to be next to impossible.  There are
> many false leads (key-text combinations which result in what seems
> like valid plaintext) and the search space becomes too large.>

> You've added the silly limiting condition that you are referring only to
> messages with two characters, which has virtually nothing to do with most
> messages sent by OTP.

No, I'm saying if the key is about half the text -- THAT ALL YOU GET TO ANALYZE.

> And, I observe if indeed you are only referring to two character messages,
> then the One Time Pad with half that number of characters, ONE character,
> amounts to the same thing as ROT-13, except where the number "13" is allowed
> to vary between 0 and 25.

Perhaps you didn't see my followup, but you still don't get it:  there
are many false positives:  decrypts that lead to semi-valid texts
WHICH YOU HAVE NO WAY TO DETERMINE VALIDITY (given no futher data).
You don't know if the key is 1 char or 2 characters, so there is a
complete set of valid decrypts which may or may not be the original
and you have no way to determine that.

marxos