feat(aztec-nr): V2 handshake registry for non interactive constrained delivery by vezenovm · Pull Request #23278 · AztecProtocol/aztec-packages

vezenovm · 2026-05-14T15:01:59Z

Fixes https://linear.app/aztec-labs/issue/F-586/handshake-registry-non-interactive-handshake-function (updates work from #22854)

Updated the handshake registry following the new spec in https://www.notion.so/aztecnetwork/Plan-Onchain-constrained-delivery-34fa1f6b0e358063b64ecc25b768c359

We store the raw handshake secret in the Handshake note
The raw secret never leaves the registry. The registry app-siloes the secret against the msg_sender and returns it from the both the handshake and its utility method for fetching the siloed secret.
The registry provides a validation method to check the app siloed secret is for a valid handshake secret
Various additional tests. Mainly making sure we reject invalid handshakes and that we never expose the raw handshake secret
Linked directly to https://linear.app/aztec-labs/issue/F-653/route-handshake-log-through-do-private-message-delivery-to-preserve to implement as a follow-up

I decided to just update the handshake registry directly rather than duplicating it with the old one. I felt if we ever wanted to go back to the old spec, we have the git history which we can reference for the old contract.

Note packing bug was revealed with this work: https://linear.app/aztec-labs/issue/F-665/note-properties-generates-incorrect-selectors-for-custom-packed-fields
I attempted to fitler notes after calling get_notes which through testing revealed itself as the incorrect way to filter notes. This should be made easier as it is a foot-gun for developers: https://linear.app/aztec-labs/issue/F-666/get-notes-and-view-notes-make-it-easy-to-filter-after-pagination

vezenovm · 2026-05-14T15:13:10Z

+    env.call_private(sender, registry.validate_handshake(sender, recipient, zero_padded_secret));
+}
+
+// Regression for the unfiltered-page bug: insert 16 unrelated handshakes (different recipients) before


This and the get_app_siloed_secret test variant are regressions for https://linear.app/aztec-labs/issue/F-666/get-notes-and-view-notes-make-it-easy-to-filter-after-pagination. I originally did not use .select on the NoteGetterOptions and filtered the result of get_notes in the utility function which caused these tests to fail. I decided to leave these tests in for now as protection. But I felt aztec-nr should make it impossible for devs to run into such a foot-gun.

nchamo

Small initial review, will continue but wanted to report early findings

…eNote, Context>, Context>, Context> and additional test

nchamo

Ok, I need to go, but I wanted to share all my thoughts so far

nchamo · 2026-05-14T19:20:11Z

    /// polluting registry state.
    #[external("private")]
-    fn non_interactive_handshake(sender: AztecAddress, recipient: AztecAddress) {
+    fn non_interactive_handshake(sender: AztecAddress, recipient: AztecAddress) -> Field {


Ok, I went through the rabbit hole. I was concerned that an attacker could front-run a handshake, which would cause our transaction to fail (it would fail since a both txs would emit the initialization nullifier). But in PrivateMutable, the initialization nullifier requires the owner's nullifier hiding key, which an attacker won't have. So you can only specify a sender that you have the keys for, which is exactly what we want: flexibility + security

On the other side, with our previous storage, anyone could have called this, since PrivateSet has no check like that. So iterating through all of them would have been very expensive

Not sure if this is 100% correct, since I'm very tired 😅 . So could we add a check to verify? I would like to test that:

An account can't set the handshake for another sender

An account can set the handshake for another sender, if the sender is added as additional scopes

Added non_interactive_handshake_rejects_other_sender and non_interactive_handshake_accepts_other_sender_in_additional_scopes.

nchamo · 2026-05-14T19:46:21Z

+    /// The raw ECDH shared-secret point `S = eph_sk * recipient_address_point`. Only this module can read it for
+    /// siloing, and it is never returned by an external function.


Only this module can read it for siloing, and it is never returned by an external function.

Nice, really like this approach!

…hake_registry_contract/src/main.nr Co-authored-by: Nicolas Chamo <nicolas@chamo.com.ar>

nchamo

Great work, we are almost there!

nchamo · 2026-05-15T15:59:46Z

-    /// Returning the latest (rather than the first) lets a sender re-initiate by issuing a fresh handshake; e.g.
-    /// if the previous secret was leaked.
+    /// Apps that receive an `app_siloed_secret` from an untrusted source (e.g. a PXE oracle) call this once to
+    /// bind that secret to the registry's stored handshake. Subsequent uses can rely on the prior-nullifier chain


"bind" feels off. I think we should stick to "validate" (and maybe rephrase the rest of the sentence with it)

nchamo · 2026-05-15T16:00:42Z

-    /// Returning the latest (rather than the first) lets a sender re-initiate by issuing a fresh handshake; e.g.
-    /// if the previous secret was leaked.
+    /// Apps that receive an `app_siloed_secret` from an untrusted source (e.g. a PXE oracle) call this once to
+    /// bind that secret to the registry's stored handshake. Subsequent uses can rely on the prior-nullifier chain


I would remove the entire "Subsequent uses can rely on the prior-nullifier chain..." text. The registry doesn't know about nullifier chains, nor should it. It just stores handshakes, and validates secrets

Co-authored-by: Nicolas Chamo <nicolas@chamo.com.ar>

…interactive

nchamo

Great work!

I'm really happy with our approach, I think this is going to be a really cool feature with minimal changes needed 🎉

AztecBot · 2026-05-15T19:10:56Z

Flakey Tests

🤖 says: This CI run detected 1 tests that failed, but were tolerated due to a .test_patterns.yml entry.

\033FLAKED\033 (8;;http://ci.aztec-labs.com/a24769b1003e89f9�a24769b1003e89f98;;�):  yarn-project/end-to-end/scripts/run_test.sh simple src/e2e_epochs/epochs_mbps.parallel.test.ts "builds multiple blocks per slot with L2 to L1 messages" (338s) (code: 0) group:e2e-p2p-epoch-flakes

BEGIN_COMMIT_OVERRIDE fix(aztec-up): fall back to no timeout when /usr/bin/timeout absent (macOS) (AztecProtocol#23310) chore: reduce compat e2e timeout (AztecProtocol#23318) feat(aztec-nr): V2 handshake registry for non interactive constrained delivery (AztecProtocol#23278) chore(aztec-nr): Public internal/utility methods self constructor (AztecProtocol#23115) END_COMMIT_OVERRIDE

vezenovm added 2 commits May 14, 2026 10:52

udpates to handshake registry

98b9784

nargo fmt

27e5a5e

vezenovm requested a review from nchamo May 14, 2026 15:01

vezenovm requested review from LeilaWang and nventuro as code owners May 14, 2026 15:02

vezenovm changed the base branch from next to merge-train/fairies May 14, 2026 15:02

clean up bug repros and link to TODO

381858d

vezenovm commented May 14, 2026

View reviewed changes

vezenovm added 2 commits May 14, 2026 11:18

fmt

fff3541

constants test update

3202ab1

nchamo reviewed May 14, 2026

View reviewed changes

nchamo self-requested a review May 14, 2026 18:13

switch to handshakes: Map<AztecAddress, Owned<PrivateMutable<Handshak…

49bfefc

…eNote, Context>, Context>, Context> and additional test

vezenovm commented May 14, 2026

View reviewed changes

Comment thread ...noir-contracts/contracts/message_discovery/handshake_registry_contract/src/handshake_note.nr Outdated

remove old comment referencing issue

d00136d

nchamo reviewed May 14, 2026

View reviewed changes

vezenovm and others added 2 commits May 14, 2026 16:42

Update noir-projects/noir-contracts/contracts/message_discovery/hands…

76c025e

…hake_registry_contract/src/main.nr Co-authored-by: Nicolas Chamo <nicolas@chamo.com.ar>

pr comments, cleanup, additional tests

75554b1

vezenovm requested a review from nchamo May 15, 2026 15:12

nchamo reviewed May 15, 2026

View reviewed changes

vezenovm and others added 4 commits May 15, 2026 14:11

Apply suggestion from @nchamo

b49b101

Co-authored-by: Nicolas Chamo <nicolas@chamo.com.ar>

comments

caf560b

,.

672e09b

Merge branch 'merge-train/fairies' into mv/v2-handshake-registry-non-…

858bc73

…interactive

vezenovm requested a review from nchamo May 15, 2026 18:41

nchamo approved these changes May 15, 2026

View reviewed changes

vezenovm enabled auto-merge (squash) May 15, 2026 18:52

vezenovm merged commit d3cf01b into merge-train/fairies May 15, 2026
14 checks passed

vezenovm deleted the mv/v2-handshake-registry-non-interactive branch May 15, 2026 19:11

AztecBot mentioned this pull request May 15, 2026

feat: merge-train/fairies #23317

Merged

		/// The raw ECDH shared-secret point `S = eph_sk * recipient_address_point`. Only this module can read it for
		/// siloing, and it is never returned by an external function.

Conversation

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

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nchamo left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

nchamo left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

nchamo left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

nchamo left a comment

Choose a reason for hiding this comment

Uh oh!

AztecBot commented May 15, 2026

Flakey Tests

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

vezenovm commented May 14, 2026 •

edited

Loading