RE: CBV

On Fri, 7 May 2004 [email protected] wrote:
> On Thu, 6 May 2004, Seth Goodman wrote:
> > On Thu, 6 May 2004, Tony Finch wrote:
> > >
> > > Note that this simple form of hash is not as safe as you might expect. I'm
> > > told by a security expert: "MD5 and SHA-1 output an intermediate value of
> > > their computation from which a hash of the original string plus an
> > > extension can be calculated. They can therefore not be used safely to
> > > calculate a MAC as h(key, string), hence HMAC." So we should use RFC2104
> > > hashes.
> >
> > This is big news if it is true.  I always thought SHA-1 was suitable for
> > HMAC's, but I would really like to know if it is not.  If what you are
> > saying is true, isn't it a problem for the existing SRS address format as
> > well?
>
> I am not aware of this weakness. I have also done my research. I would
> like to see a proper reference for it.

I think there's a bit of miscommunication here. HMAC (described in RFC
2104, using either MD5 or SHA1) is the correct way to create the hashes.
Creating a MAC by simply taking the MD5 or SHA1 of a secret concatenated
with the public data to be authenticated is not good enough.

RFC 2104 has references that describe why this is the case.

-- 
Tony Finch  <[email protected]>  http://dotat.at/