Re: md5 collision generator

Matthew D. Fuller wrote:
> On Wed, Nov 16, 2005 at 10:29:09AM -0500 I heard the voice of
> Tom Lane, and lo! it spake thus:
>
>>The existence of this algorithm is disturbing, since it implies that
>>MD5 is weaker than people thought,
>
> It occurs to me that, controlling everything that would be poking into
> that part of the database, it would be possible to store the password
> with several DIFFERENT hash algorithms, which would save us in the
> future from any of them being easily crackable (or even ALL of them,
> unless you can somehow create a collision across them all
> simultaneously). It seems that even with 2 or 3 weak hashes, that
> might be safer long-term than with just 1 strong hash. I s'pose it
> would add a little cost to the connection-establishing process...

We really should be using an HMAC instead of a simple hash anyway. I
don't believe, even given the attacks available on MD5 and SHA1, that
anyone has theorized or demonstrated any weakening of HMACs based on
these hash algorithms (someone please correct me if they know otherwise).

Joe