On Wed, Mar 30, 2016 at 7:02 AM, <[email protected]> wrote:

> 2) I am fairly certain that any ideas like this should not concern either
> of the NTS-related documents. Whether/how to use any samples whose
> authenticity & integrity can *not* be verified by NTS should be out of
> scope for us.
>
> I disagree. The whole purpose of NTS is to protect timing information, so
at the very least the document should clearly and precisely explain the
bootstrapping problem (what to do if the client has no idea what time it is
upon startup) and the implications of using unauthenticate time.  Punting
on this means that the draft ignores a central issue in secure time
synchronization.

The language in Section 9.2 of the current draft could be strengthened.

Re:bootstrapping:

I suggest that NTS should encourage users to write the current time to a
persistent file that is available upon reboot.  This should be overwritten
on every clock update.    Then, when the client reboots, it can check that
the expiry time of the certificates is no earlier than the last time time
written to the presisent file.  This limits the impact of attackers that
use old compromised certificates to break the security of NTS.

Re: using unauthenticated time samples.

Leaving this out of scope gives uses just enough rope to hang themselves.
The draft should explain the risks and implication of using unathenticated
time. As I discussed in my previous email, doing this allows even a
*boosttrapped* client to be attacked.  I would even suggest that the draft
use MUST to forbid the use of unauthenticated samples.  If this is already
in the draft and I have missed it, please accept my apologies.

Thanks,
Sharon



-- 
Sharon Goldberg
Computer Science, Boston University
http://www.cs.bu.edu/~goldbe
_______________________________________________
TICTOC mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/tictoc

Reply via email to