[WIMSE] New I-D: OASNT, attested action authorization tokens

[Arun Thallapelly <[e-mail]>]

Hello,

I have submitted draft-thallapelly-oasnt-00, which defines a compact JWS in
which a hardware-bound device key attests that a specific human authorized
a specific action, with the human-readable disclosure they were shown bound
into the signature.

It is implemented: a C++ minter on device, a JavaScript verifier, and a
public adversarial corpus. The appendix carries test vectors generated from
that implementation, so an independent verifier can be validated against it.

I would value the list's view on two questions:

   1. The three digests are separate top-level claims ("adg", "dsp", "rqf")
   rather than one nested object. I chose that because "rqf" is optional and
   its absence is load-bearing during verification. Is that the right trade?
   2. How should this relate to the human-authorization-binding work
   already circulating here? I would rather align with that work than
   duplicate it.

A constraint worth flagging for anyone specifying in this area: the iOS
Secure Enclave supports P-256 only, so a specification mandating Ed25519
for human-authorization evidence cannot be satisfied by a hardware-bound
key on that platform.

Thanks,
Arun Thallapelly
OmniArx.ai