refactor!: collapse create_session proof type into session x create by kilianglas · Pull Request #866 · worldcoin/world-id-protocol

kilianglas · 2026-07-14T12:03:14Z

Removes ProofType::CreateSession and redesigns the request model: proof_type shrinks to {uniqueness, session} and session_id becomes a three-state field. Creating a session was always create-and-prove (a session proof in the same response), so it collapses into proof_type: "session" + session_id: "create" with no capability change. Hence, there are only two proof types now, Uniqueness Proofs and Session Proofs. Both can create a session, only session proofs can be bound to an existing session. Therefore, sessions can either be bound to a nullifier, or can be started without any binding. Note, that Uniquess Proofs can still be requested without creating or binding a session.

Stacked on #862.

Request model

New SessionRef type carried by ProofRequest.session_id:

wire value	state
absent / `null`	`SessionRef::None`
`"create"`	`SessionRef::Create` — mint a fresh session, prove it in the same response
`"session_<hex>"`	`SessionRef::Existing(SessionId)`

Note: The SessionRef is not in the signed part of the request.

Validation matrix (validate_proof_type):

	`null`	`"create"`	`"session_<hex>"`
`uniqueness`	✅ plain	✅ create bound session	❌
`session`	❌	✅ (was `create_session`)	✅

action remains forbidden for proof_type: "session" (both sub-states).

Different combinations have different use cases:

uniqueness, null: Standalone uniqueness proof that does not require session.
uniqueness, "create": Uniqueness proof that requires bound session for later session proof.
session, "create": Create new standalone session to prove human continuity without uniqueness guarantees. Used e.g., in the upcoming DeepFace proof type.
session, "session_<hex>": Session proof for existing session. This is used for any session_id, uniqueness bound or not.

Changes to OPRF nodes

To make this work, the auth module of the ORPF nodes had to be slightly changed. OPRF queries with prefix 0x01 (used to obtain the session seed), can now originate from Uniqueness Proof RP requests. Those have an action field iin the signed data. Hence, in order to verify the RP Request signature, the OPRF request needs to carry this action, which differs from action = oprf_seed which is used for the query. We introduce a new field called rp_signature_verification on NullifierOprfRequestAuthV1 struct.

WIP-101 RPs

Important: Combined uniqueness + create session requests are not yet supported. This will be implemented in a follow-up PR.

…ects them

…oofs

…-session-binding

…uniqueness-session-binding-contracts

…hub.com:worldcoin/world-id-protocol into kilianglas/uniqueness-session-binding-contracts

Replace the dedicated create-session proof type with a three-state session reference so request semantics are explicit on the wire. Co-authored-by: Cursor <cursoragent@cursor.com>

Carry the RP-signed uniqueness action separately from the session OPRF seed so nodes can safely authorize atomic session creation. Co-authored-by: Cursor <cursoragent@cursor.com>

Accept create and existing session references on uniqueness requests, mint sessions through the authenticator, and validate the corresponding response semantics. Co-authored-by: Cursor <cursoragent@cursor.com>

Exercise the signed-action create flow end to end and generate verifier fixtures from a session minted by the uniqueness request. Co-authored-by: Cursor <cursoragent@cursor.com>

Document create, existing, and unbound uniqueness modes together with the signed-action OPRF flow and deployment order. Co-authored-by: Cursor <cursoragent@cursor.com>

dkales · 2026-07-17T07:45:43Z

    /// If omitted, the request is strictly treated as a [`ProofType::Uniqueness`] request.
    /// Session creation and session proving must opt in explicitly.
    #[serde(default)]
    pub proof_type: ProofType,


Old create_session requests will now fail for sure though.

Yes. @paolodamico is this an issue?

Zero is the circuit's "no session" sentinel, satisfiable by any World ID regardless of mt_index, so a session record left at zero could be satisfied by an arbitrary World ID with a patched authenticator. verifyWithSession already rejected it; apply the same check to verifySession. V2 only — older implementations are immutable. Reported via HackerOne #3912490. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

…ilianglas/collapse-create-session

V2 is released and immutable, so the new entry point and the session id validation move to a new version instead of extending V2 in place. - IWorldIDVerifierV3 declares verifyWithSession + InvalidSessionId - WorldIDVerifierV3 implements verifyWithSession and overrides verifySession to reject sessionId == 0 - WorldIDVerifierV2.sol is restored byte-identical to main; InvalidAction stays declared inline there and V3 inherits it - V2 tests return to their original set (regenerated fixture values only); the binding and zero-session cases move to WorldIDVerifierV3Test - test-utils deploys V3 behind the proxy; e2e and the fixture tool retarget Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

…ilianglas/collapse-create-session # Conflicts: # contracts/src/core/interfaces/IWorldIDVerifierV3.sol

The regenerated fixture values were unnecessary: all six V2 tests are negative (four revert on InvalidAction before any proof math, two expect ProofInvalid), so they pass unchanged against the refreshed mocks. Leaves the PR with no V2 footprint at all. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

…ilianglas/collapse-create-session

The verifySession zero-session check is a separate security fix and moves to its own PR against main. V3 here adds only the new entry point. Depends on that PR landing for the session-proof hardening. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

…ilianglas/collapse-create-session # Conflicts: # contracts/src/core/interfaces/IWorldIDVerifierV3.sol

The base branch was changed.

…reate-session # Conflicts: # contracts/test/core/WorldIDVerifierV3Test.t.sol # crates/authenticator/src/prove.rs # crates/core/tests/generate_proof.rs # crates/primitives/src/request/mod.rs # crates/test-utils/src/anvil.rs # docs/world-id-4-specs/README.md

The merge left main's session-bound section alongside the create-and-bind one, so the tool no longer compiled: session_id is now SessionRef, not Option<SessionId>. Removes the stale block and the nullifier clone it needed. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Keeps the discriminants aligned with the action prefixes: 0x00 for a uniqueness action, 0x02 for a session action. 0x01 is skipped because it prefixes the session oprf_seed rather than a proof flow. Reinstates the reservation from #711 for future signed request payloads, and pins it with a test so the values cannot renumber silently. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

0xForerunner · 2026-08-06T19:07:20Z

@kilianglas Why do we disallow uniqueness + session_<hex>?

0xForerunner · 2026-08-06T19:22:26Z

It's worth noting that this PR allows oprf nodes to link the initial uniqueness action and the session seed. I don't really think that's a problem in practice, just wanted to flag it.

0xForerunner · 2026-08-06T19:29:38Z


+        let rp_signature_verification = match (proof_request.proof_type, proof_request.action) {
+            (ProofType::Uniqueness, Some(action)) => {
+                Some(RpSignatureVerification::UniquenessAction { action })


The signed message doesn't include proof_type or session_id, which means these fields can be changed and the signature is still valid. We probably want to fix this.

This is a valid more general question. The proof type is implicitly encoded in the action (action with 0x00 prefix is present -> uniqueness, no action -> session). The session id field is currently not signed over by the RP, and there might be ways to abuse this (need to think about this). Since this is pre-existing behavior, though, I don't think this should be addressed in this PR.

kilianglas · 2026-08-07T08:13:58Z

@kilianglas Why do we disallow uniqueness + session_<hex>?

We thought it might be an anonymity footgun if we allow RPs to bind uniqueness proofs to existing sessions.

kilianglas · 2026-08-07T08:19:24Z

It's worth noting that this PR allows oprf nodes to link the initial uniqueness action and the session seed. I don't really think that's a problem in practice, just wanted to flag it.

Yes, that's true. Didn't think this was an issue, especially since r is cached anyway by the authenticator, session proofs don't need the oprf_seed round trip in most cases. Can you think of any scenario where this might be an issue?

kilianglas added 14 commits July 10, 2026 10:27

feat: allow binding uniqueness proofs to an existing session

6de6065

chore: minor cosmetic changes

82bb6ae

docs: bound proofs verifiable via verifyProofAndSignals, verify() rej…

cee2156

…ects them

feat: add verifyWithSession verifier entry point for session-bound pr…

5440e41

…oofs

feat: generate and verify bound-uniqueness solidity fixture

e07919a

test: add bound uniqueness fixtures and verifyWithSession tests

91ca3cd

docs: document verifyWithSession entry point

de729fd

refactor: move verifyWithSession to new IWorldIDVerifierV2 interface

2382247

Merge remote-tracking branch 'origin/main' into kilianglas/uniqueness…

24f9cb0

…-session-binding

Merge branch 'kilianglas/uniqueness-session-binding' into kilianglas/…

6c9a289

…uniqueness-session-binding-contracts

chore: cleanup

50777bc

Merge branch 'kilianglas/uniqueness-session-binding-contracts' of git…

7895737

…hub.com:worldcoin/world-id-protocol into kilianglas/uniqueness-session-binding-contracts

chore: drop unrelated README reformatting

8d475b1

docs: reword session binding verifier requirement

74d15fe

kilianglas requested review from a team, 0xThemis, dkales, fabian1409 and philsippl as code owners July 14, 2026 12:03

kilianglas mentioned this pull request Jul 14, 2026

feat: verify RP signature over signed action on session seed queries #865

Closed

3 tasks

kilianglas and others added 5 commits July 14, 2026 15:18

refactor!: model session creation through SessionRef

b594803

Replace the dedicated create-session proof type with a three-state session reference so request semantics are explicit on the wire. Co-authored-by: Cursor <cursoragent@cursor.com>

feat(node): authorize signed session-seed queries

a54cb50

Carry the RP-signed uniqueness action separately from the session OPRF seed so nodes can safely authorize atomic session creation. Co-authored-by: Cursor <cursoragent@cursor.com>

feat: create sessions from uniqueness proofs

f4b147b

Accept create and existing session references on uniqueness requests, mint sessions through the authenticator, and validate the corresponding response semantics. Co-authored-by: Cursor <cursoragent@cursor.com>

test: cover atomic uniqueness session creation

28dafed

Exercise the signed-action create flow end to end and generate verifier fixtures from a session minted by the uniqueness request. Co-authored-by: Cursor <cursoragent@cursor.com>

docs: describe uniqueness session creation

f651855

Document create, existing, and unbound uniqueness modes together with the signed-action OPRF flow and deployment order. Co-authored-by: Cursor <cursoragent@cursor.com>

kilianglas force-pushed the kilianglas/collapse-create-session branch from 5fee2df to f651855 Compare July 14, 2026 13:25

kilianglas added 4 commits July 14, 2026 15:45

chore: remove unnecessary tests

b7c1977

feat: rp_signature_verification field

816e9cf

refactor: single RpSignatureVerifcaton erro

8986872

chore: update comment

6db473a

docs: polish readme

25f454c

0xThemis approved these changes Jul 17, 2026

View reviewed changes

dkales approved these changes Jul 17, 2026

View reviewed changes

kilianglas added 2 commits July 21, 2026 12:54

refactor: remove unnecessary unreachable!

828b5ec

fix: remove unused imports in generate-solidity-fixtures

2d7d228

kilianglas force-pushed the kilianglas/collapse-create-session branch from faf2043 to 2d7d228 Compare July 21, 2026 12:52

kilianglas and others added 3 commits August 3, 2026 21:41

feat: remove uniqueness bindign to existing session id

c379ea9

Merge branch 'kilianglas/uniqueness-session-binding-contracts' into k…

70af89a

…ilianglas/collapse-create-session

0xThemis previously approved these changes Aug 4, 2026

View reviewed changes

dkales previously approved these changes Aug 4, 2026

View reviewed changes

kilianglas and others added 6 commits August 4, 2026 12:02

Merge branch 'kilianglas/uniqueness-session-binding-contracts' into k…

db9023a

…ilianglas/collapse-create-session # Conflicts: # contracts/src/core/interfaces/IWorldIDVerifierV3.sol

Merge branch 'kilianglas/uniqueness-session-binding-contracts' into k…

5c3a870

…ilianglas/collapse-create-session

Merge branch 'kilianglas/uniqueness-session-binding-contracts' into k…

43a9c85

…ilianglas/collapse-create-session # Conflicts: # contracts/src/core/interfaces/IWorldIDVerifierV3.sol

Base automatically changed from kilianglas/uniqueness-session-binding-contracts to main August 5, 2026 15:32

kilianglas and others added 2 commits August 6, 2026 11:15

kilianglas enabled auto-merge (squash) August 6, 2026 09:34

0xThemis approved these changes Aug 6, 2026

View reviewed changes

0xForerunner requested changes Aug 6, 2026

View reviewed changes

Navigation Menu

Uh oh!

refactor!: collapse create_session proof type into session x create - #866

refactor!: collapse create_session proof type into session x create#866
kilianglas wants to merge 44 commits into
mainfrom
kilianglas/collapse-create-session

kilianglas commented Jul 14, 2026 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

Uh oh!

dkales Jul 17, 2026

Uh oh!

kilianglas Jul 18, 2026

Uh oh!

0xForerunner commented Aug 6, 2026

Uh oh!

0xForerunner commented Aug 6, 2026

Uh oh!

0xForerunner Aug 6, 2026

Uh oh!

kilianglas Aug 7, 2026

Uh oh!

kilianglas commented Aug 7, 2026

Uh oh!

kilianglas commented Aug 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Conversation

kilianglas commented Jul 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Request model

Changes to OPRF nodes

WIP-101 RPs

Uh oh!

Uh oh!

Uh oh!

Uh oh!

dkales Jul 17, 2026

Choose a reason for hiding this comment

Uh oh!

kilianglas Jul 18, 2026

Choose a reason for hiding this comment

Uh oh!

0xForerunner commented Aug 6, 2026

Uh oh!

0xForerunner commented Aug 6, 2026

Uh oh!

0xForerunner Aug 6, 2026

Choose a reason for hiding this comment

Uh oh!

kilianglas Aug 7, 2026

Choose a reason for hiding this comment

Uh oh!

kilianglas commented Aug 7, 2026

Uh oh!

kilianglas commented Aug 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

kilianglas commented Jul 14, 2026 •

edited

Loading