On Oct 21, 2013, at 5:11 AM, Dan York <[email protected]> wrote:
> Good document. In a quick read I naturally have to react to slide 23
> (Trusting DNS) and also slide 11 (Certificate Authority).  On slide 23,
> you say "Sorry, can't trust this yet.", but what happens as we get more
> DNSSEC deployed?  We're already seeing increased validation within caching
> resolvers and some measurements are showing around 8% of all DNS queries
> coming from resolvers that perform validation.  We're seeing steady growth
> in the number of DNSSEC-signed domains.  I know there are those who are
> skeptical about DNSSEC deployment, but I'm definitely seeing real
> growth... and see a number of trends pointing to that only continuing.

DNSSEC has two huge issues and one huge potential (untapped) advantage.

Issue #1:  Validation at the wrong place.  Recursive resolver validation is of 
only small (but non-zero) utility: It allows countering packet-injection (NSA 
and their foreign competitors), which can otherwise be used to promote a 
man-on-the-side to a full man-in-the-middle.  But the recursive resolver itself 
is proven untrustworthy, so its not the point which should conduct validation.

And if clients are to validate in a mandatory way, they need to be able to get 
DNSSEC over an alternate means, as many clients can't get DNSSEC over the wire 
on UDP 53, many others can't get DNSSEC from the recursive resolver, and the 
combination (can't get either from the wire OR recursive resolver) is high 
enough that its in the "break 1%" category, which is generally regarded as 
unacceptable.


Issue #2:  The big threat against DNS, and the big reason you can't trust it, 
is that the major high-profile attacks have been attacks on the registrar.  
DNSSEC does no good, because if you can social engineer the registrar, you can 
also get new DS records accepted.



Untapped advantage:  Multiple paths of trust.  If, say, TLS was modified to 
have multiple hostnames, eg
https://www.example.com/www.example.ru/

where the resulting certificate is fetched through DNSSEC, and BOTH domains 
must agree, in order for the connection to proceed.  This provides a very 
unique property: it requires that all paths-of-trust be compromised by the same 
cooperating entity.  That is

. -> .com -> example.com
. -> .ru -> example.ru

both must be compromised by the same entity.  

"I don't trust .com, I don't trust .ru, but I trust they won't collude against 
me" is a very interesting property that is unique to protocols built on top of 
DNSSEC-as-a-CA.


--
Nicholas Weaver                  it is a tale, told by an idiot,
[email protected]                full of sound and fury,
510-666-2903                                 .signifying nothing
PGP: http://www1.icsi.berkeley.edu/~nweaver/data/nweaver_pub.asc

Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail

_______________________________________________
perpass mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/perpass

Reply via email to