openssh unix dev - Mar 2011 - Indicating context when asking the ssh-agent to use a key

This e-mail is about thinking through some possible extensions to the
ssh-agent communications protocol.


When the ssh-agent receives a request to use one of the keys it holds,
it gets no context information from the requesting system about what
the key operation is to be used for.

My own typical workflow (as a user who actively monitors and confirms
the use of my keys by the ssh-agent) is to just correlate things by
time.  e.g. "i just did action X, so i expect key Y to be used right
around now, so i'll say OK".

If there was a way to communicate the context of the use to the agent,
so that the agent could relay that to the user in whatever notification
or confirmation it provides, it would seem like a Good Thing.

If there was a way to do that with some measures of cryptographic
reliability (e.g. so that a malicious client couldn't say "please make
this signature for X" when it was actually intending to be used for Y),
it would be even better.  I'm not sure i understand how that could
happen, though i'd be happy to consider proposals/suggestions.

I suspect this would require at least an extension to the ssh-agent
protocol, but i'm not sure where or how that would be done.


Any thoughts on this?


i just opened a bug report about this here, if anyone wants to
contribute proposed patches/protocol suggestions:

 https://bugzilla.mindrot.org/show_bug.cgi?id=agent-context


I'm happy to have a bigger-picture discussion here on the list, though.
 Is this a bad idea?  a nice idea but unimplementable for some reason?
is this already possible somehow?

Regards,

	--dkg

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 1030 bytes
Desc: OpenPGP digital signature
URL:
<http://lists.mindrot.org/pipermail/openssh-unix-dev/attachments/20110307/4e026f85/attachment.bin>