Re: New hash function: avalanche

From: Paul Crowley (paul@JUNKCATCHER.ciphergoth.org)
Date: 02/20/03 

From: Paul Crowley <paul@JUNKCATCHER.ciphergoth.org>
Date: Thu, 20 Feb 2003 02:25:06 GMT

NOSPAMperlmodules@lunkwill.org (Jason) writes:
> Thusly, I have constructed a new hash function which is guaranteed to
> change *all* output bits for every bit of change in the input.

This is possible only for a hash function that takes at most two
possible inputs. If there are three or more possible inputs (say A, B
and C are three such), then consider bit i of the output. If it is
different for A and B, and different for B and C, then it must be the
same for A and C, therefore changing A to C does not change bit i. 

-- 
  __  Paul Crowley
\/ o\ sig@paul.ciphergoth.org
/\__/ http://www.ciphergoth.org/