Re: [Netconf] Call Home for Notifications

[Kent Watsen <[e-mail]>]

On 12/12/12 3:24 PM, "Phil Shafer" <[e-mail]> wrote:

>Juergen Schoenwaelder writes:
>>Back im ISMS days, some security folks argued that even then this is
>>not the same. The reasoning, as far as I understood it and recall it,
>>which may be completely wrong, was that in the normal case, the client
>>has a clear idea to which host it wants to connect and then the client
>>can check the discovered host key against what was expected. If you
>>turn this around, the 'client' gets presented with a host connecting
>>and since the initiative come from the remote entity, the client has a
>>harder time to decide whether the host is meaningful to engage into a
>>conversation with. This may sound like an artificial philosophical
>>argument, perhaps it is, perhaps it is just my poor recollection and
>>paraphrasing of it.
>
>Does having the device pre-configured with the hostkey of the manager
>help?  It then knows that the incoming connection is from someone that
>it's supposed to be talking with.


Assuming you meant host key of the *device*, absolutely, but
draft-kwatsen-reverse-ssh takes it a step further by enabling trust in the
host-key to be learned, per it being cryptographically signed with
something the manager trusts.

Juergon, I'm familiar with the need for TLS clients to verify a
certificate's subject matter matches expectations, as it is prudent to
validate the identity your peer.  In the case of device-initiated
connection, there is still a desire to validate its identity, though it
may take on other forms including serial-number, hash(serial-number),
ip-address, and/or fqdn.  Ideally the manager has some precognition of one
of these, but this may not always be possible.  This is why
draft-kwatsen-reverse-ssh introduced the ability for a CA to sign the
device's credentials, thus enabling a manager to extend its trust for the
CA to the device, even if never seeing its identity before.

FWIW, neither the IETF-SSH nor SAAG raised any security concerns when
discussing draft-kwatsen-reverse-ssh, though this may be due to them being
too preoccupied with protocol-purity and never getting to think about it ;)


K.