feat(confirmations): scan EIP-712 address fields via phishing controller

[wzrdk3lly]

Summary

Ticket: PSAFE-441

MetaMask validates eth_signTypedData_v3 and eth_signTypedData_v4 requests with PPOM, which inspects the address fields of a signature. PPOM's threat data is refreshed on a delay, so an address that has been flagged recently can still be reported as benign, and it does not cover every chain or protocol.

Separately, the real-time per-address scan (/address/evm/scan) is only applied to a few fields today: the token verifyingContract, permit spender, and delegation delegate. Addresses carried in any other field — recipients, makers, tokens, custom protocol fields — receive no real-time scan at all.

This change wires a schema-driven extractor (added to @metamask/phishing-controller in MetaMask/core#9875) that returns up to 10 distinct address-typed values found in a typed-data message, including nested structs and arrays, matched by declared type rather than field name. When PPOM has not already flagged the request, those addresses are passed to PhishingController.scanAddress. Results reuse the existing address cache, so fields already scanned elsewhere are not requested again. A confirmation alert surfaces any flagged address using the existing alert template, naming the flagged field and address. If some addresses could not be checked because the cap, depth limit, or work budget was reached, a separate caution is shown.

Changes

• app/lib/address-scanning/scan-unvalidated-signature.ts (new) — extracts up to 10 address-typed fields and scans each via PhishingController.scanAddress after PPOM; excludes signer and zero address
• app/components/Views/confirmations/hooks/alerts/useSignatureAddressAlerts.ts (new) — hook that reads scan results and returns alerts for malicious/warning addresses (naming the field) plus a caution when the walk was incomplete
• app/lib/ppom/ppom-util.ts — calls scanUnvalidatedSignatureAddresses after PPOM for v3/v4 requests when PPOM has not flagged the request
• app/components/Views/confirmations/constants/alerts.ts — adds SignatureAddressScanIncomplete, SignatureAddressTrustSignalMalicious, SignatureAddressTrustSignalWarning to AlertKeys
• app/components/Views/confirmations/hooks/alerts/useConfirmationAlerts.ts — wires useSignatureAddressAlerts into useSignatureAlerts()
• app/components/Views/confirmations/hooks/metrics/useConfirmationAlertMetrics.ts — adds metric entries for the three new alert keys
• locales/languages/en.json — adds alert_system.signature_address_scan.* strings

Dependencies

Blocked on MetaMask/core#9875. The package.json dependency on @metamask/phishing-controller must be bumped to the version that exports extractSignatureAddresses before this can merge.

References

• Core PR (extractor): feat(phishing-controller): export extractSignatureAddresses for EIP-712 typed-data scanning core#9875
• Extension (parallel client wiring): feat(confirmations): scan EIP-712 address fields via phishing controller metamask-extension#45521
• Prior implementation with local extractor copy (superseded): feat(confirmations): scan address fields in signature requests #34428

Test plan

• Open a v3/v4 typed-data signature on a supported chain with a known-malicious address in a non-from field — confirm a Danger alert appears naming the field and shortened address
• Open a permit-type signature — confirm no duplicate spender alert
• Trigger a message with more than 10 distinct address fields — confirm the scan-incomplete caution appears
• Confirm no alert regression on standard transaction confirmations
• Verify metrics fire for the three new alert keys

Screenshots/Recordings

Both recordings sign the same typed-data request — the message carries a flagged address in a non-permit field.

Before

Baseline build. The flagged address raises no alert on the confirmation screen.

simulator.screen.before.mov

After

With this change, the same request shows a danger alert on the "Interacting with" row naming the field and address.

Simulator.screen.after.mov

---

Note

Medium Risk
Touches signature confirmation security alerts and PPOM post-validation scanning. Behavior is additive and gated, but incorrect extraction or alert mapping could hide or over-flag threats.

Overview
Adds real-time phishing scans for address-typed fields in eth_signTypedData_v3/v4 messages, covering nested structs and custom fields that PPOM or permit-only scans miss.

After PPOM, if the result is not already malicious or warning, scanUnvalidatedSignatureAddresses extracts addresses (excluding the signer) via extractSignatureAddresses and scans them through PhishingController. Confirmation UI then shows danger/warning alerts naming the field and address, plus a caution when the extractor hits its cap.

Gated on security alerts / Blockaid being enabled. Personal sign and typed-data v1 are skipped. Metrics keys are added for the three new alerts.

^{Reviewed by Cursor Bugbot for commit 098fd99. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

CLA Signature Action: All authors have signed the CLA. You may need to manually re-run the blocking PR check if it doesn't pass in a few minutes.

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Reviewed by Cursor Bugbot for commit e9ad9bd. Configure here.}

[matthewwalsh0]

Can we use something more explicit like useSignatureTrustSignalAlerts?

[wzrdk3lly]

Great point! I applied this change

[matthewwalsh0]

Minor, hook was renamed but not file itself.