[webauthn] new commits pushed by agl

The following commits were just pushed by agl to https://github.com/w3c/webauthn:

* The DPK is stored on the authenticator.

The text said that the DPK was stored on the client device, but the
client device is the device that the browser is running on, not the
authenticator. That _might_ be the same device, but it's unclear.
Clarify that the DPKs come from the authenticator.
  by Adam Langley
https://github.com/w3c/webauthn/commit/d92bad2a3f5925b4dd4d0cb39544393e7f3a491b

* Provide attestation controls.

Mirror the attestation controls for user credentials into the DPK
extension.
  by Adam Langley
https://github.com/w3c/webauthn/commit/6d45aba6a5088cac743f85c33fec48dec60bd010

* Pull out DPK attestation rules and add signature prefix.

This change adds a section about calculating DPK attestations and
references that each time rather than duplicating the rules. It also
adds a prefix to the signed messages to ensure that DPK and user
credential attestations clearly cannot be confused.
  by Adam Langley
https://github.com/w3c/webauthn/commit/eb598ff6a4d08ce35262ee2de3537a8989809cf9

* Reflow CDDL to avoid a scroll bar.
  by Adam Langley
https://github.com/w3c/webauthn/commit/b7289e1686669ae00f65653ef608c594a504513f

* Have the DPK sign over everything.

Signing over too little is a common problem in protocols and the DPK
wasn't signing over very much. The problem is that the signature was
within the extension itself and thus couldn't easily cover the
authenticator data.

This change puts the DPK signature next to the normal signature in an
assertion, transforming that field into a CBOR array. That's fine for
assertions, but registration doesn't have such a signature output. Thus
this change drops exercising the DPK during registration:

Firstly, this is in line with the user credential which doesn't sign
during registration unless "self" attestation is used. (And "self"
attestation can be use with the DPK if desired.)

Secondly, adding an extra signature output for registration is awkward.
Putting it in the user credential attestation statement is awkward and
conflicts when user-agents replace attestation statements. Adding a new
CTAP field is possible but seems excessive since, as noted, there has
never been a signature by the user credential.
  by Adam Langley
https://github.com/w3c/webauthn/commit/dcfb39270f989b30dce6772d03e58c4549902afd

* Note that CTAP2 CBOR is required in DPK.
  by Adam Langley
https://github.com/w3c/webauthn/commit/cbb6b5d050a0c7647521da9ae9c6858c878d0f17



-- 
Sent via github-notify-ml as configured in https://github.com/w3c/github-notify-ml-config

Received on Thursday, 19 May 2022 22:58:03 UTC