Re: [core] 🔔 Confirming adoption of draft-hoeglund-core-oscore-key-limits-02 as a CoRE WG document

[Christian Amsüss <[e-mail]>]

Hello John, KUDOS authors,

On Mon, Nov 15, 2021 at 10:22:46AM +0000, John Mattsson wrote:
> I think it would be good to discuss if the KUDOS rekeying mechanism
> should/could be used to also update the identifiers. KUDOS resets the
> sequence numbers. I have not thought about this in any detail or that
> it is something we should do, I just suggest that we discuss it.

The two are related but at different layers; any solution combining them
would need to operate on both. (KUDOS sending unprotected nonces, new
KIDs would need to be negotiated in encrypted data).

It may help to see them independent initially:

* KUDOS allows using new key material from a preexisting context, with
  sequence numbers starting at 0 again, but keeps the same KIDs.

* KID switchovers could be announced using inner options -- "Please
  address me as ${my_new_sender_ID} henceforth".

  Keeping the master key, whoever changes their KID needs to be aware of
  all KIDs previously used on that shared context. In particular, the ID
  needs to stay unused until the peer has acknowledged (eg. by using the
  new ID) that it didn't try to switch to that ID at the same time.

The requirement to know history (lest a sender key get drived a second
time) is what makes KID changes convenient to do at KUDOS time (when the
list of previously used KIDs is empty again).

BR
c

-- 
To use raw power is to make yourself infinitely vulnerable to greater powers.
  -- Bene Gesserit axiom