A revised "FFT" hash function.
From: Tom St Denis (tomstdenis_at_iahu.ca)
Date: 09/30/03
- Next message: Bryan Olson: "Re: controversial paper"
- Previous message: George Ou: "Re: controversial paper"
- Next in thread: Mads Rasmussen: "Re: A revised "FFT" hash function."
- Reply: Mads Rasmussen: "Re: A revised "FFT" hash function."
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Date: Mon, 29 Sep 2003 23:15:01 GMT
First off disclaimer, don't use this design [at least right away] in
anything production. You've been warned.
The "niche" of this hash function is hardware specifically but it can also
work in embedded software [requires ~200 bytes of ram]. It uses a radix-2
FFT network for diffusion and a simple 8x8 as the nonlinear part.
In hardware the design would be very fast as the critical path is rather
short [the FFT network has a path of maybe 5*3 == 15 xor gates] and it would
be
compact as the main compnents are a 4x256 [fastest, or 9x8 for smallest]
SRAM, a bunch of XOR gates and a series of 96 4x4 ROMs The design is a
request of Mads Rasmussen so don't flame me please :-)
The reference source [which isn't terribly optimized] is at
http://iahu.ca:8080/src/ref.c and test vectors are at the end of the source
code. Please don't assume anything and avoid using this design in
production systems. It's an "interesting" design perhaps at best. As for a
paper on the design I doubt I will write one but Mads may want to. I wrote
the code for him.
Tom
- Next message: Bryan Olson: "Re: controversial paper"
- Previous message: George Ou: "Re: controversial paper"
- Next in thread: Mads Rasmussen: "Re: A revised "FFT" hash function."
- Reply: Mads Rasmussen: "Re: A revised "FFT" hash function."
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Relevant Pages
|
|