Re: [pkix] Validating Certs w/out reliable source of Time

[Tom Ritter <[e-mail]>]

On Mon, 8 Oct 2018 at 14:13, Dr. Pala <[e-mail]> wrote:

> Hi Panos, all,
> thanks for the info. It seems nobody has a good story around it - the
> onboarding provides some obvious paths, but it does not provide really a
> good story around it and it is very prone to implementation errors (it
> seems more like giving up in having a good answer / system when you do not
> trust the network itself - which is the case I am trying to cover).
>
> Although I totally agree with the difficulty around providing a solution,
> I am a bit worried about devices keeping logs/audit traces and then follow
> up on them at a later time - especially without providing guidance about
> what is a trusted source of time... :D I would expect many devices not to
> really check the validity of certificates after they have been "used"
> already.
>
> In my specific use-case (which is not a generic case), I am leaning toward
> building a signed time service w/ a simple challenge-response mechanism
> that can be proxy and verified by the device... since we already have
> domain-specific trust anchors deployed, we might leverage those also for
> this use-case.
>

If you use OCSP in a challenge-response mode with nonces - you could get
this within established TLS/PKIX standards. Although OCSP
challenge/response (as opposed to stapling) is falling out favor; and OCSP
nonces stopped being used over a decade ago. But I don't think the code is
removed from tools; just uncommonly used.

Caveats being; of course, the uptime of your time/OCSP server; difficulty
of rotating that server's certificate, what to do when you don't get a
response....

-tom