Re: [TLS] Server time

[Florian Weimer <[e-mail]>]

On 04/06/2015 02:45 PM, Peter Gutmann wrote:
> Tom Ritter <[e-mail]> writes:
> 
>> Is it better or worse than relying on a completely unauthenticated protocol
>> to get the time?
> 
> If you really need a secure time source you either use NTPv4 security

NIST has published a protocol for NTPv4 key negotiation:

  <http://www.nist.gov/pml/div688/grp40/auth-ntp.cfm>

Yes, it's a manual protocol, due to the lack of better alternatives.

> or run your own stratum 1.

That's still subject to spoofing.  One solution is to obtain
cryptographically secured time from the national time source (which
exists in most countries, I assume), with a baked-in (CA) certificate.
If it's wrong, it's still correct because it's the official time.

The Network Time Security effort more or less explicitly excludes
importing time from the outside (over the Internet or otherwise) because
it's requirements RFC, RFC 7384, explicitly states this is a
non-requirement in section 3.2.10.  (The RFC also has requirements which
seem quite impossible to implement.)

The current NTS protocol proposal uses CMS over UDP for key negotiation,
but has some issues.  I don't know yet if they can be persuaded to adopt
DTLS (maybe lack of server-side state is absolutely critical to them).

-- 
Florian Weimer / Red Hat Product Security