Re: How do we get the whole world to upgrade to DNSSEC capable resolvers?

[Eric Rescorla <[e-mail]>]

At Tue, 12 Aug 2008 15:45:57 -0400 (EDT),
Dean Anderson wrote:
> On Mon, 11 Aug 2008, David Conrad wrote:
> > 
> > > Only SSL can protect you here.
> > 
> > As Dan Kaminsky points out: "SSL certs themselves are dependent on the  
> > DNS".
> 
> Kaminsky is wrong. SSL uses DNS to obtain an IP address to connect to a
> server, and then expects the server to produce a certificate, which the
> client verifies. Spoofing DNS does not enable the attacker to obtain the
> private key to the valid certificate.  The DNS domain information placed
> in a certificate merely allows the client to determine before
> verification if it got the certificate it asked for. However the wrong
> or fake certificate won't verify unless there is a fault in the
> Certification Authority (this happened with MS some years ago). But a
> spoofed DNS name going to a wrong server results in a failure to verify
> the certificate. So SSL is NOT "dependent on DNS".

Based on my brief skim of Kaminsky's presentation, I think his
argument is that because CAs often use e-mail answerback as a
certificate validation mechanism, a compromise of DNS can be used to
obtain a fake SSL/TLS certificate. Of course, there's nothing stopping
CAs from requesting more validation (in fact, I believe that's what EV
certificates already do), in which case there would not be a DNS-based
threat.

That said, it's not clear to me that the situation is significantly
better with DNSSEC.  Both certificates and signed zone signing keys
are attestations that a given public key corresponds to the owner of a
given domain name/group of names. The non-cryptographic question is
how the signer of the credential determines whether a given candidate
public key is correct. But this question is more or less the same for
DNS servers and CAs, because in both cases there is generally not a
cryptographic credential, but just an authoritative contact point,
which often reduces to an email address.

-Ekr
 

--
to unsubscribe send a message to [e-mail] with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/namedroppers/>