Hi, Shane:

Shane Kerr wrote:
> I'm sorry if it has already been discussed, but has there been any work
> done on using TLS for AXFR/IXFR?

There was a mini-thread back in June 2016, but I think that was it:

https://mailarchive.ietf.org/arch/msg/dns-privacy/eNHQx5KT3MzqjFObqgDRswCpxi4

> It seems like it should be relatively straightforward, compared to the
> stub-to-resolver and resolver-to-authority links. While it does not
> seem as big of a problem either, obviously somebody cares about hiding
> the contents of zones or everybody wouldn't block zone transfers,
> right? 

I think this was explicitly mentioned at least in passing in "DNS
Privacy Considerations" (RFC 7626), in §2.1 "The Alleged Public Nature
of DNS Data".

DPRIVE has done some work to foil traffic analysis attacks with the
EDNS0 padding option, so it makes sense to me to also encrypt zone
contents during distribution, because zone contents are the exact data
you would need to execute a traffic analysis attack.

DNSEXT defined NSEC3 on the explicit rationale of preventing zone
enumeration. Implementing TLS for zone transfers would address the
collection of zone content by pervasive monitoring.

> There are still some issues to consider, such as what the interaction of
> TSIG and TLS certificates means, as well as what method a master can
> use to signal TLS support (this seems desirable to me, although not
> necessary).

The easiest thing to do is probably to make it all a matter of explicit
configuration in server configuration files, for which system
administrators have a well-known solution (configuration management).

Zone transfers already require explicit configuration of several
parameters, anyway, and adding TLS will require a bunch more (e.g.,
permitted TLS versions, permitted cipher suites, certificate validation,
etc).

> Does this seem like something worth working on? Or is it a distraction
> from the resolver-to-authority work?

I think if resolver-to-authority confidentiality is implemented, the
next logical plaintext leak to plug would be zone transfers, and if the
full charter were implemented ("…or provide end-to-end confidentiality
of DNS transactions"), then the lack of zone transfer confidentiality
would be a glaring omission.

> Note also that it might be worthwhile building a new zone transfer
> protocol that can perform better in areas where AXFR and IXFR don't
> work well today (unnecessary data in IXFR of signed zones, inefficiency
> for synchronizing lots of zones, automatic fallback to full zone
> transfer on IXFR failure, and so on). That's not really something for
> DPRIVE, of course, but adding TLS to the protocol could be rolled into
> such an activity.

But if an enhanced zone transfer (EXFR?) type were implemented, it seems
like it would be nice to be able to run it over plaintext port 53, so I
don't see why it should be tied to the transport.

(BTW, one thing I wonder about is, for operators who *don't* use
AXFR/IXFR on port 53 to distribute their zones, does the transport they
use support confidentiality equivalent to TLS/SSH?)

-- 
Robert Edmonds

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

Reply via email to