Re: Added hashes.

Joseph Ashwood <ashwood@msn.com> wrote:
> "Ian Stirling" <root@mauve.demon.co.uk> wrote in message
> news:439b6724$0$63085$ed2e19e4@ptn-nntp-reader04.plus.net...
>> Isn't it more secure to hash a file by hashing it twice with two seperate
>> functions, and then xoring the result?
>> Is this more secure than both, or only as secure as the weakest link?
>
> First I will agree with Hadstate, all you've done is create a new hash
> function from the old, it will still have a limited lifespan, and will be
> only half the speed of the originals (currently this is breaking down
> because of multi-core CPUs, but we'll just live with it for now). With that
> said, the proposed solution is not bad as a stopgap measure, but will not
> suffice long term.

Thanks both of you.
I'll disagree slightly with the 'half speed' comment - as in many cases
the limiting factor is disk speed, but of course it may be an issue.

I should have been clearer about my question.

If I've got a block of data to hash, and want a hash function that will
not let anyone else compute a block of different data with the same hash
function, then if each of the seperate hash functions are broken (it's possible
to compute a block to yield an arbitrary output) then isn't the task of
computing a block that will give the correct combined hash a very different
one from breaking each seperately?

Or once each has been broken, is this a fairly trivial step, with better
security only till all of the component hashes have been broken.

> Right now we are in a watershed time for hash functions. Up until just a few
> years ago, not much was understood about them, the MD series was written by
> one man (Rivest), and most of the others are very similar. Recently though
> with the CryptoNessie process and the breaking seemingly everything we have
> gained noticable new insight, and our new insight has lead to some
> interesting findings. Currently there are two hashes that are interesting
> for the simple fact that they bear no resemblance to the MD/SHA series,
> these are Panama and Whirlpool. While there was an announcement some time
> ago that Panama had been broken, I am unclear on the exact details and I
> don't remember them ever being published. Both of these are Wide Trail
> Strategy designs, while the MD/SHA series are Feistel designs. This recent
> knowledge is in it's infancy and the next 10 years should be an interesting
> time for hashes.

Especially for those that have to pick between competing hashes, without
detailed knowledge or understanding of why they will break.
Received on Fri Dec 23 20:09:32 2005