FROST co-signing for shielded ZEC: specification draft v0.1, review wanted
The first draft of the open specification from ZCG grant #490 is public: docs/spec.md. It covers how two devices of one user share a FROST 2-of-3 account for shielded ZEC in the Ironwood pool, so that a co-signer from one wallet can work with a coordinator from another:
- transport:
ur:zcash-frostmessages, as animated multi-part QR codes or as text; - pairing: an X-Wing key agreement with commit-reveal, a 6-digit code compared on both screens, then sealed messages;
- key generation: the FROST DKG, 2 of 3, with
skagreed by commit and reveal and the keys derived as ZIP 2005 specifies; - signing: three messages, the co-signer’s commitments first;
- what the co-signer checks before it signs, and what it shows the user;
- member backups, in the format of the backup ZIP draft (zcash/zips#1432);
- test vectors.
frost-zec implements it, and OpenPGP Wallet 2.16 uses it with a test co-signer. The testnet and mainnet spends will be in the Milestone 1 report, in this thread.
Where a second pair of eyes would help most:
- Pairing (§3). Is a 6-digit code over X-Wing enough here? A pairing has no forward secrecy beyond device compromise; is that acceptable for v1?
- The end of key generation (§4). The coordinator has the account as soon as it holds the co-signer’s keys, but the co-signer finishes members 2 and 3 only from the last message. In one of our own testnet runs, the account was funded before that message arrived, and those coins can never move. The wallet now asks the user to confirm that the co-signer shows the address. Should the protocol add a last sealed acknowledgement from the co-signer instead?
- Signing in two QR codes instead of three (§9). The co-signer would commit after seeing the coordinator’s commitments and the message. Is that safe with FROST’s binding factors, given that it could no longer check the aggregated signature before its user approves?
- Signer checks (§6). Any PCZT the co-signer should refuse and does not. The current cases are in tests/tamper.rs.
Comments here or as issues on frost-zec are welcome. Version 1.0 will follow the review, in Milestone 2.