Notes: Janfred Rieckers
Agenda remains unbashed.
Alan presenting.
Interop testing shows problems.
Proposal: Publish 7170bis as TEAPv1 with current behavior, write a
TEAPv2 spec with more clear spec and do interop testing before
publishing
Joe (as participant): Removing things that don't work is probably best
thing to do. Sounds like the EMSK-derivation seems straight-forward, but
compound-MAC may need more consideration, may be a security issue
Alan: you have MSK and EMSK, you derive some from that which you put in
Crypto-Binding-TLV. Works with one MSK and EMSK, but on the second round
there is no explicit spec and implementations did it differently
Joe: Definitely have to talk more about that. You still have some
compound binding? It just doesn't go all the way?
Alan: Its inner tunnel binding vs TTLS session. TTLS will derive the
CHAP/MSCHAP challenges from TLS-PRF, those are tied to the TLS tunnel,
this is not true for EAP-TLS in TEAP. TTLS has the same problem if you
run EAP-TLS inside TTLS
Joe: To clarify: You don't have just no EMSK crypto binding, you have no
crypto binding, right?
Alan: Yes.
Alan: Other thing: Currently for basic password the spec says to derive
an MSK of all zeros, maybe you could derive it from the outer TLS-PRF
Joe: I don't know if the binding does anything
Alan: Means that there is no one in the middle meddling
Joe: Cryptobinding is to bind the outer key to inner key. With password
we don't have an inner key.
Joe: Original spec had obviously to many options
Alan: Tried to explain how to do this multiple MSK/EMSK derivation in
the current doc, feedback was that it's still unclear and complicated.
We could strip all that for TEAPv1, and for TEAPv2 define a better
method
Joe: Chaining in general is a problem, would be good to move quickly.
Alan: Suggestion: we strip TEAPv1 down to the currently interoperable
functionality (MSCHAPv2, MSCHAPv2 followed by MSCHAPv2, EAP-TLS, EAP-TLS
followed by MSCHAPv2)
Question to the AD from the Chairs: Document is currently already
requested for publishing, what is the process?
Alan: It's in RFC editor queue, but it's broken enough that we have to
change pages of text
Paul: Let's move it back to the working group
Dan Garcia-Carrillo presenting.
Question from Dan:
Joe (as participant): Would depend on the use cases, is fast reconnect
necessary?
Peter: Ask mailing list for further comments
Janfred: What are your use cases? RFC4017 talks about requirements for
EAP methods for WLAN
Dan: Primary use case is IoT, will also look at it if it is useful
Janfred Rieckers presenting.
Janfred: Will change the name, soon. There are running code and working
interop between different implementations. Discussion on binary formats
and cryptographic binding. Cryptographic binding needs to be discussed
more. Will talk to W3C and FIDO Aliiance about specifying a general
clientData format. Can we use the TLS negotiated hash algorithm. Hash
algorithm is negotiated with the hash algorithm and can often be
extrated.
Hannes: The changes to binary etc, moves this away from FIDO,
interesting decision.
Janfred: Decision if SHA-256 should be hardcoded.
Quynh: Unclear if the hash function itself is accesible.
Hannes and Joe: Discussion on identity authentication, which is more
often done in HTTPS than EAP.
Hannes: Your method was inspired by FIDO but is no logner FIDO. Should
describe differences from FIDO
Tim: You cannot expect credentials to be available for different use
cases.
Tim: Everything is defined in WebAuth. If you call if WebAuth, you need
to follow the WebAuth rules including user presence.
Vadim: Is the network another WebAuth app?
Alex: Should maybe resurrect TEAP pack something...
Alan: How do you provision over unauthenticated?
Hannes: Provisioning is ignored and many of the existing solutions are
bad. Thanks for working on provisioning.
Joe pesenting (as WG Chair)
We want to do the adoption call for this together with the recharter, so
that if the recharter is accepted, we can immediately adopt this
document.
Hannes: Could you add "for example" or similar to reference privacy pass
protocol and tokens, maybe reword it to a wider variety, and not locked
to this specific protocol/building blocks
Paul (AD): Is there a draft?
Joe: Yes, there is already a -01
Joe: We could also rework the charter and remove elements we don't need
any more
Hannes: Would it be good to also add provisioning to the recharter?
Joe:
Hannes: Revolutionary assumption in PPT .... ,
maybe can also helpful for provisioning. In line with some current
architecture to have a different interface for connection and
provisioning.
Tiru presenting.
Hannes: Great work, definitely needed.
Quynh: also support this.
Relay from chat: Could you also implement these updates to AKA' to
wpa_supplicant? Helps with adoption.
Tiru: I'm more a 3GPP person, so don't know much about that.