Prohibition of alg=none not quite enough

[Nat Sakimura]

Comments

Since we are trying to integrity protect the authorization response, the use of alg=none needs to be prohibited. Although the first bullet in Clause 3 Client Metadata states

The algorithm none is not allowed.

it is not using normative language and is too weak.

Also, it would be a good idea to include the algorithm check in 2.4 Processing rules so that conformance test specifically checks for it.

Proposal

Amend the sentence above to read as:

The algorithm none (“alg”:”none”) MUST NOT be used.

Also, insert the following above 2.4-2 stating that if alg=none then the message MUST be rejected.

‌

[Nat Sakimura]

Brian Campbel wrote in https://lists.openid.net/pipermail/openid-specs-fapi/2022-July/002637.html

It is prohibited in the authorization_signed_response_alg client metadata but I agree that explicitly prohibiting it in the client side processing rules would be good to add. Probably in the authorization_signing_alg_values_supported AS metadata too.    

‌

[Torsten Lodderstedt]

+1

[Nat Sakimura]

Accepted.

Brian will make a PR.

[Brian Campbell]

PR #351