Review · October 26, 2026
Additional Authentication Method Reference Values Summary
This draft adds nine new values to the IANA amr (Authentication Methods References) registry and rewrites the definition of one existing value, tel. That's the whole spec. There are no new claims, no new protocol messages, and no new processing rules.
The Problem
RFC 8176 created the amr registry in 2017. Since then, passkeys, push approval, authenticator apps, and silent network authentication have gone mainstream. None of them has its own registered value. An identity provider that wants to tell a relying party "the user approved a push notification in Okta Verify" has no standard string for it.
The draft says the values are general purpose. The one use case it names is the Skyfire KYAPay token.
How It Fits In
This is a registry update. It depends on RFC 8176 and, indirectly, on JWT (RFC 7519) and OpenID Connect Core, which defines the amr claim. It doesn't touch OAuth itself. It adds no roles and no new relationships between client, authorization server, and resource server.
It's in charter in the loose sense that the OAuth WG has handled JWT claim and registry work before. The document is marked as updating RFC 8176 because it changes the meaning of tel.
It belongs to the Skyfire family of drafts alongside the KYAPay token, using KYAPay tokens, ID verification, and AML methods. Revision -01 moved the inp and vid values out of this draft and into the ID verification draft. That was the right call, since those describe identity proofing rather than authentication.
So does this need a working group document at all? Mostly no. The AMR registry uses Specification Required. Any stable public spec can register new values through the designated experts. The tel change is the exception. RFC 8176 lists the IETF as change controller for tel, so changing its meaning needs an IETF document.
What It Defines
New values:
app: authenticator app (Okta Verify, Google Authenticator, Microsoft Authenticator)bg: background authentication (silent network auth of a phone number, device recognition, geolocation)email: code or link sent to emailcode: code provided to and entered by the userurl: URL provided to and clicked by the userpush: push notification to a mobile phonefacliv: facial recognition with a liveness checksqa: security question answerspasskey: passkey
Updated value:
tel: broadened from "confirmation by telephone call" (call back) to any call to a registered number, including one that just reads out a code for the user to type somewhere else.
Each definition is one line. There's no further text describing what the AS asserts when it emits a value.
Significant Interoperability Concerns
Many of these values overlap with ones already in the registry. RFC 8176 already defines otp, kba, face, hwk, swk, sms, geo, rba, and user. Compare:
sqa(security questions) vs.kba(knowledge-based authentication). In practice these are the same thing.faclivvs.face. Liveness is a real distinction, but the draft never says whether an AS should sendfaclivalone orfaceplusfacliv.passkeyvs.hwk/swkplususer. A synced passkey is a software key. A device-bound one is a hardware key. The new value hides that distinction.bgvs.geoandrba.bgexplicitly includes geolocation, which already has its own value.appvs.otpandpush. Okta Verify can do either one. Does an AS sendapp,otp,push, or all three?
The draft never addresses any of this. Two identity providers will describe the same login with different arrays. A relying party writing policy like "require phishing-resistant auth" has to know every combination each IdP might send.
code, url, and email describe how a secret is delivered, not the authenticator. These sit in a different dimension from the rest of the registry. A code sent by email could be ["email"], ["email", "code"], or ["email", "otp"]. A magic link could be ["email"] or ["email", "url"]. code alone doesn't even say which channel delivered it. The registry already struggles with mixed granularity. This makes it worse.
The tel change silently weakens an existing value. Today, a relying party that sees tel can assume the call itself was the confirmation. After this change, tel can also mean "we read you a code over the phone." That's a meaningfully weaker assurance. Relying parties that already make decisions on tel get no signal that the meaning shifted. A new value for "code delivered by voice," next to sms, would avoid the problem.
bg is too broad to act on. Silent network authentication through a mobile carrier and "we recognized the device cookie" carry very different assurance. Putting both under one string means the RP can't distinguish them.
Mechanisms That Could Be Removed
sqa: duplicateskba. The draft itself notes that NIST no longer accepts security questions. Registering a new name for a deprecated method doesn't help anyone.codeandurl: delivery details rather than authentication methods. Drop them, or define them only as qualifiers with explicit combination rules.bg: either split it into specific values like silent network auth and device recognition, or drop it in favor ofrbaandgeo.- Everything except the
telchange could go through the normal designated expert process without a working group document.
passkey, push, and app are the clearest wins. Those are what implementers are actually asking for. Even those need a sentence each on how they relate to the existing hwk/swk/otp values.
Writing Quality
Clean and correct. This uses Mike Jones's standard registry template, and the IANA section is complete. The problem is that it's too thin. Each value gets a single descriptive line with no guidance on when to emit it, how it combines with existing values, or how it differs from its nearest RFC 8176 neighbor. RFC 8176 is short too, but its values were mostly orthogonal. These aren't, so they need more text.
The security considerations add one useful sentence about security questions. They say nothing about the tel semantics change, which is the one part of this draft with real security impact.
Authors
Datatracker wasn't reachable from this session. This section is based on the drafts in the repo and on published history.
- Ankit Agarwal (Skyfire Systems): New to the IETF. Appears only on the Skyfire family of drafts, where he is very active right now.
- Michael B. Jones (Self-Issued Consulting): Author of dozens of RFCs, including JWT, JWS, JWK, and RFC 8176 itself, which this draft updates. Co-author on every Skyfire draft and many other active drafts.
- Nash Ali (Experian): New to the IETF. Appears only on Skyfire drafts.
- Srinivasa Thumma (Akamai): New to the IETF. Appears on this draft, the ID verification draft, and the using-KYAPay-tokens draft, all added in recent revisions.
Bottom Line
The gap is real. Passkeys and push approval should have registered amr values, and the industry will benefit from them. The draft adds values without reconciling them against the existing registry, so the same login ends up with several correct encodings. Feedback should push for three things. Cut or merge the values that duplicate RFC 8176 (sqa, bg, and arguably facliv). Define how the remaining values combine with existing ones. Register voice-delivered codes as a new value instead of redefining tel. Once tel is out of the picture, the authors can ask whether the rest could simply go to the designated experts.