EMU @ IETF119

Notes: Janfred Rieckers

Administrative

Agenda remains unbashed.

WG Items

EAP-AKA-FS

Jari Arkko: There are no technical issues, good to go to waiting for
IESG processing.

Bootstrapped TLS

Going with the downref for RFC 8773 as the timeline to making RFC 8773
(bis) a Proposed Standard within the TLS WG is unclear.

Charter Update

Porposed update posted on Github, to incorporate new work done, namely
FIDO and EDHOC.
Looking to add some deliverables in the charter.

EDHOC -> Authentication for IoT/constrained environments

• Dan Harkins: What's the client credential?
  • Göran: COSE-specified format for credential, could be X.509,
    could be Cbor Web Token, could be Raw Public Key. EDHOC supports
    what can be identified by a CBOR header map registered in IANA.

EAP-FIDO -> Authentication using FIDO2

• Tim Cappalli: Struggling with this because FIDO2 is based on signing
  over the web origin; struggling to understand how this fits to EAP.
• Chair: Can we hold this for the FIDO presentation slot. Most of the
  participants are not WebAuthn, FIDO, etc. experts.

eap.arpa -> Add a .arpa domain to support EAP methods. Should we add it
to the charter?

Future WG Items

eap.arpa

Alan DeKok presenting.

EAP-NOOB specified eap-noob@arpa as username for provisioning.
The .arpa domain is a special domain, not routable, under IAB control.
Existing implementations just deal with it as if it was any other
domain.

.arpa is under IAB control, so we could register eap.arpa with them
and then use an IANA registry for usernames (left of @), with a common
space (for things like dpp@eap.arpa or portal@eap.arpa) and a vendor
specific space.

• Chairs: Is currently in WG adoption call, probably not long until
  WGLC

EAP-EDHOC

Göran presenting.

There was an addoption call between IETF 118 and now, committed to get
reviews from EAP and EDHOC view.

Waiting for the revised charter to finally adopt this.

There is implementation effort and interest from the industry.

There is an Hackathon in Paris in May (invitation open to everybody)
https://parishackathon.lakewg.org/

• Chairs: Next step is to get the charter updated, so we can formally
  adopt this.

EAP-FIDO

Janfred Rieckers presenting.

One string to rule them all: "anonymous@\" This is the
answer to Tim\'s earlier question. If not configured to use discoverable
credentials, then the user would need to provide a \'username\'.

Janfred happy to have collaborators who are familiar with FIDO. Would
ideally like to ensure the user doesn't have to keep pressing the FIDO
button all the time.

Carsten's student, so would like to use CBOR, but EAP uses TLVs; not a
fan, (rather use CBOR maps) but can adjust if necessary. Still working
with Carsten on the right CBOR model.

Would be good to have link/liaison to the FIDO alliance.

Chair: We did reach out to John Bradley as far as someone with FIDO
experience to look at the document; he said yes, but hasn't happened
yet.

Alan: The operators have been complaining that the TLS provisioning is
really painful. This stuff doesn't work cross platform. It would be good
to get this.

Tim: There isn't a web identity available to the EAP supplicant. You
can't connect the identities to the FIDO system. I have lots of
experience with FIDO, this can't connect and work.
Janfred: more and more devices have a FIDO implementation within, so if
the OS offers some way of interacting, i.e. by CTAP2, we can use that to
trigger silent authentication.
Tim: Platforms themselves don't use CTAP2; the tokens cannot be used
silently. Would have to completely blow up the spec.

Alan: Similar problems with PEAP and TTLS and resumption; too much load
on campuses, resolved by resumption; same thing might apply with
passkeys; if you could reuse credentials from a passkey that's been
authenticated, that would be really helpful to places such as campuses.
The fallback problem is not different than what we have now, people pay
a lot of money to a lot of companies to get EAP working, often it
doesn't

Tim: the problem is about bootstraping; that's a challenge; connecting a
web-tied credential to EAP. Same problem with bootstrapping
certificates. Many actions that the user can take to stop FIDO from
working (i.e. deleting the credential, unplugging authenticator).
Nothing wrong technically with the proposal, may just not work in the
way envisioned in practice

Janfred: Silent authentication is in the CTAP2 standard, so why doesn't
that work?
Tim: Passkeys are defined in WebAuthn; there's some unfamiliarity with
the division between WebAuthn and CTAP2, with WebAuthn handling the Web
stuff, but CTAP2 handling everything else; passkey requires human
interaction to use per its standard.

Alexander Clouter: The problem with interactive vs non-interactive is
that we could bring back the PAC (Protected Access Credential) in TEAP
as a token or extend the session token. You can enroll temporarily using
passkey to get a long-lived session token. TEAP did this with PAC.
Janfred: really sceptical, want to get rid of certificates.

Heikki: Would it be useful to have split spec for EAP and one for
provisioning?
Janfred: This wouldn't work well, this is basically already done, the
TLS layer is already a separate spec (RFC 5216) and this spec just
specifies how to do FIDO over that TLS layer.

Alan: sounds like the hardest part of this is getting access to the
credentials (esp. platform authenticators). Less worried about
disappearing credentials, people mess EAP-TLS based configuration up
today already.
OS Vendors should just allow access to passkeys from the supplicant as
well.

Chair: Would like to understand what the constraints are to know how
well this has the potential to work.

Tim: biggest fear this ends up a wpa_supplicant on Linux only
implementation. What would this even look like on Linux which doesn't
even have a FIDO2/WebAuthn platform authenticator? So how would this
work?

Janfred: This is valuable feedback; have a lot of material to go through
for the next revision. If you or someone else would be available to
discuss, that would be helpful.

Non-WG Items

EAP-AKA-PQC

Aritra Banerjee presenting

Jari: thanks for this, we need something like this. Happy to help on
this. Not all the details are clear.
AB: need to do updates to get to -01 draft and replace with PQ
algorithms

Chair: since this stuff is new, it would be good to get some formal
analysis of this done; the new integration of algorithms and the hybrid
design. Not necessarily a requirement, but would be helpful.
AB: Absolutely

Multiple Pre-Shared Keys (EAP-MPSK)

Speaker is Lei Yan.

Chair: Has the draft been posted to the list?
Yan: Yes, it has.
Chair: We should see if there is interest on the list; doesn't seem like
a lot of interest at this point.
Chair: Please try to follow up on the list to find more interest in this
work.