Posted by trlists on 04/22/05 13:45

On 21 Apr 2005 Greg Donald wrote:

> > Same thing with MD5, it
> > is just one way, it can't be reversed.
>
> MD5 collisions were found last year:
> http://cryptography.hyperlink.cz/md5/MD5_collisions.pdf
>
> Just a matter of time/cpu power.

I don't think that's right. Collisions allow certain kinds of
cryptographic attacks against things like MD5-based signatures but that
is not at all the same as being able to simply determine the original
message content from the digest. Rather, they allow you to substitute
the original message with a different one which generates the same MD5
hash. This may or may not be useful as an attack, depending on how MD5
is being used.

--
Tom

[Back to original message]