Repository navigation
perf(profile-sync-controller): Replace JS AES implementation with cryptography package - #10621
Merged
Merged
Conversation
FrederikBolding
marked this pull request as ready for review
September 30, 2026 13:15
FrederikBolding
marked this pull request as draft
September 30, 2026 13:29
FrederikBolding
force-pushed
the
fb/profile-sync-encryption
branch
from
October 1, 2026 08:18
f34700a to
f7600fc
Compare
FrederikBolding
marked this pull request as ready for review
October 1, 2026 09:17
Contributor
|
@metamaskbot publish-preview |
Contributor
|
Preview builds have been published. Learn how to use preview builds in other projects. Expand for full list of packages and versions. |
mathieuartu
reviewed
Oct 1, 2026
mathieuartu
left a comment
mathieuartu
left a comment
Contributor
There was a problem hiding this comment.
Tested manually on extension, features relying on it work without issues. Left one question, would feel more comfortable if somebody more crypto-litterate than me would also have a look, but overall looks good
Comment on lines
-222
to
-225
| const nonce = randomBytes(ALGORITHM_NONCE_SIZE); | ||
|
|
||
| async #encrypt( | ||
| plaintext: Uint8Array<ArrayBuffer>, | ||
| key: Uint8Array<ArrayBuffer>, | ||
| ): Promise<Uint8Array<ArrayBuffer>> { | ||
| // Encrypt and prepend nonce. | ||
| const ciphertext = gcm(key, nonce).encrypt(plaintext); |
Contributor
There was a problem hiding this comment.
I'm really not an expert on those matters, so please bear with me here. Genuine question: wasn't ALGORITHM_NONCE_SIZE of importance here?
Member
Author
There was a problem hiding this comment.
@metamask/cryptography uses a 12-byte nonce/IV as well! It is equivalent, just generated within the encryption function: https://github.com/MetaMask/core/blob/main/packages/cryptography/src/aes-gcm.ts#L5
Contributor
There was a problem hiding this comment.
I wonder if we could have exported this constant? E.g we could have re-used it while decoding the bytes on the other part of the code?
So we avoid having drifiting values for ALGORITHM_NONCE_SIZE and AES_GCM_IV_LENGTH
Contributor
There was a problem hiding this comment.
Ah, makes sense, thanks! But then, if it changes one day in @metamask/cryptography, wouldn't we start having problems here:
async #decrypt(
ciphertextAndNonce: Uint8Array<ArrayBuffer>,
key: Uint8Array<ArrayBuffer>,
): Promise<Uint8Array<ArrayBuffer>> {
// Create buffers of nonce and ciphertext.
const nonce = ciphertextAndNonce.slice(0, ALGORITHM_NONCE_SIZE);
const ciphertext = ciphertextAndNonce.slice(
...The fact that rely both on a "silent" external constant AND a local one is making me a bit uncomfortable. WDYT?
Member
Author
There was a problem hiding this comment.
That's fair, we can export it and re-use!
Member
Author
There was a problem hiding this comment.
Though tbf this shouldn't change, the defacto standard is 12 bytes.
Contributor
There was a problem hiding this comment.
yep that's right, I feel like it still helps with the code here (especially for the encoding/decoding part)
Contributor
There was a problem hiding this comment.
I agree and am inclined to approve as-is. But for the sake of future readers, this asymmetry should probably be fixed in a follow up PR.
…yptography package
FrederikBolding
force-pushed
the
fb/profile-sync-encryption
branch
from
October 1, 2026 11:43
9e30663 to
0a3cbc4
Compare
mathieuartu
approved these changes
Oct 1, 2026
FrederikBolding
changed the title
ccharly
approved these changes
Oct 1, 2026
ccharly
left a comment
ccharly
left a comment
Contributor
There was a problem hiding this comment.
LGTM (left 1 nit about wording, but that's really minor and not necessarily "incorrect")
Comment on lines
+233
to
+244
| ciphertextAndNonce: Uint8Array<ArrayBuffer>, | ||
| key: Uint8Array<ArrayBuffer>, | ||
| ): Promise<Uint8Array<ArrayBuffer>> { | ||
| // Create buffers of nonce and ciphertext. | ||
| const nonce = ciphertextAndNonce.slice(0, ALGORITHM_NONCE_SIZE); | ||
| const nonce = ciphertextAndNonce.slice(0, IV_LENGTH); | ||
| const ciphertext = ciphertextAndNonce.slice( | ||
| ALGORITHM_NONCE_SIZE, | ||
| IV_LENGTH, | ||
| ciphertextAndNonce.length, | ||
| ); | ||
|
|
||
| // Decrypt and return result. | ||
| return gcm(key, nonce).decrypt(ciphertext); | ||
| return decrypt(key, nonce, ciphertext); |
Contributor
There was a problem hiding this comment.
Nit: But I think we can use the right wording now, I'd go with Iv/iv instead of Nonce/nonce
Member
Author
There was a problem hiding this comment.
They are interchangeable tbh, do you feel strongly about this?
Contributor
There was a problem hiding this comment.
Not that much 😄 I usually see IV rather than nonce in the context of AES, but that's not incorrect neither
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Explanation
Replace
@noble/ciphersAES implementation with@metamask/cryptographyin encryption utilities used byprofile-sync-controller. As a result of this I also had to change some of the internal types to the narrowerUint8Array<ArrayBuffer>type.References
https://consensyssoftware.atlassian.net/browse/WPC-1331
Checklist
Note
Medium Risk
Changes the AES-GCM implementation backing client-side profile sync encryption while keeping the same payload format; wrong crypto behavior would break decrypt compatibility across clients.
Overview
Profile sync encryption now uses
@metamask/cryptography/aes-gcminstead of@noble/ciphers, with@metamask/cryptographyadded as a dependency and@noble/ciphersremoved. The on-disk layout is unchanged: scrypt-derived keys and IV + ciphertext still use a 12-byte IV (viaIV_LENGTH), but#encrypt/#decryptare async and byte handling goes through@metamask/utils(stringToBytes,concatBytes,getErrorMessage).Internal types are tightened to
Uint8Array<ArrayBuffer>in the encryption cache, KDF paths, andNativeScrypt.createSHA256Hashnow returns hex without a0xprefix (remove0x(bytesToHex(...))). Tests expect generic WebCrypto-style decrypt failures instead of noble’s invalid ghash tag message.Monorepo wiring adds
@metamask/cryptographyto TS project references,tsconfig.packages.jsonpaths, JestmoduleNameMapper, README dependency graph, and changelog entry.Reviewed by Cursor Bugbot for commit 8e2bd03. Bugbot is set up for automated code reviews on this repo. Configure here.