# 2-of-3 Multisig: The Freeze Stopped the Co-Signer, Not the Coins

By Pawnie (@pawnie_ on X). Web3. Published 2026-11-20. Canonical page: https://www.w3match.com/blog/2-of-3-multisig-who-controls-keys-unocoin

A BitGo freeze stops BitGo's own co-signing, not a transfer signed with the other two keys. Custody models compared, with the Unocoin case as the example.

A freeze is a promise about one party's behaviour. It is not a lock on the coins. The Unocoin and BitGo incident shows the difference, and the two companies now describe the mechanics in much the same way.

## What the parties say, and what the chain shows

BitGo says the affected wallet was a 2-of-3 self-custody setup: Unocoin held the user key and the backup key, BitGo held the third. BitGo says Unocoin froze its enterprise on 6 November at 11:29 UTC, and that the Ethereum transactions were signed with the user and backup keys and broadcast outside BitGo. [BitGo statement, 13 November](https://www.bitgo.com/resources/blog/unocoin-security-incident/)

Unocoin says (13 November) the freeze "functioned as intended" and (14 November) that BitGo's infrastructure was not compromised and BitGo did not sign. [Unocoin update](https://blog.unocoin.com/security-incident-update/)

Bitquery's on-chain analysis, independent of both, found 98 transfers between 9 and 11 November on Ethereum, all signed by the user key and the backup key. The backup key had never sent a transaction before 9 November. [Bitquery](https://bitquery.io/investigations/unocoin-hack) That shows which keys signed. It does not show who used them or how they were obtained.

## What Unocoin says about its keys

Unocoin says its backup keys were compromised. It says they were encrypted, printed on paper about eight years ago and kept in physical storage, and that the decryption passphrase existed in its compromised infrastructure, which it says "is not fully solving the puzzle". How the backup keys were obtained is not stated. [Unocoin update](https://blog.unocoin.com/security-incident-update/)

Unocoin estimates the loss at about USD 200,000 from 3 of more than 20 hot wallets and says it will cover it. As of 14 November that is a commitment.

Bitquery separately traced about USD 182,000 from one Ethereum wallet and 39,167 XRP (about USD 42,700) from a second account. These are partial traces with different scopes. We do not add them to each other or to Unocoin's estimate.

## What a freeze can and cannot do

BitGo's support documentation says a freeze prevents BitGo from co-signing, and that someone holding both the user and backup keys can sign without BitGo. [BitGo support](https://support.bitgo.com/support/solutions/articles/158000454794-security-incident-response) Its developer documentation says recovery transactions "bypass any policies you have with BitGo". [BitGo recovery docs](https://developers.bitgo.com/docs/wallets-recover) A wallet freeze blocks outgoing transactions through BitGo for a set period. [BitGo freeze docs](https://developers.bitgo.com/docs/wallets-manage-freeze)

So the freeze worked as designed, and the design does not cover that path. The path exists so a customer is not locked out if the provider disappears. It is equally open to whoever assembles two keys.

"2-of-3" means two independent parties only if the two non-provider keys are really separate: different people, systems and passphrase storage. Two keys with one holder act as one.

## Custody models compared

| Custody model                                                 | Who holds keys                                               | What a freeze/co-signer can stop              | What it cannot stop                               |
| ------------------------------------------------------------- | ------------------------------------------------------------ | --------------------------------------------- | ------------------------------------------------- |
| Single-key self-custody                                       | You                                                          | Nothing, no provider involved                 | Anyone with the seed moving funds                 |
| 2-of-3, you hold two keys (the Unocoin setup)                 | You: user and backup. Provider: third                        | The provider's own co-signing                 | A transfer signed with your two keys              |
| 2-of-3, backup with a separate recovery service (our reading) | You: user. Recovery service: backup. Provider: third         | The provider's co-signing                     | A transfer signed by you and the recovery service |
| Full custody                                                  | Provider holds all three, split among its staff (BitGo docs) | Every path is the provider's                  | Nothing you control                               |
| MPC/TSS                                                       | Shares across several parties; full key never created        | Depends on which share holders can still sign | A threshold of compromised shares                 |
| Social-login shares (SSS, per Web3Auth docs)                  | OAuth node network, device, backup; key rebuilt on device    | Pausing the login service                     | Keys already exposed                              |

BitGo argues multisig shows signers on-chain more clearly than MPC. That is a vendor position, and BitGo also sells MPC wallets. Keplr's exact configuration was not disclosed.

## Other ways keys and credentials get exposed

Chainstack says an attacker entered a development environment on 26 October through a third-party component and was cut off on 31 October. It treats node credentials as exposed, though it says it has no evidence they were read. [Chainstack report](https://chainstack.com/security-incident-nov-2026/)

Socket's analysis found a malicious @ethersphere/bee-js 13.3.0 on npm on 9 November, which npm removed after about 75 minutes. [Socket](https://badge.socket.dev/blog/bee-js-compromise) Victims and losses: not stated.

Web3Auth and Keplr say a Web3Auth flaw exposed keys of some Keplr wallets made with Google or Apple login. The Block's on-chain review found about USD 5.6 million in Web3Auth recovery addresses on 8 November. That is held funds, not a loss figure. [The Block](https://www.theblock.co/news/defi/2026-11-08-web3auth-says-it-frontran-hackers-by-moving-5-6-million-out-of-exposed-keplr-wallets-436029) On 13 November Web3Auth said the funds were still held. [Web3Auth update](https://web3auth.io/updates/security-incident-update.html) The theft total is not stated.

## Questions to ask your custody provider

- Which keys exist, who holds each, and do any two share a holder or a system?
- Where is each passphrase stored, and is that system separate from the key backup?
- What exactly does a freeze stop, and what does it not stop?
- Which routes bypass our policies, such as whitelists and approvals?
- Is activity on rarely used backup keys monitored?
- After a suspected key compromise, how are keys rotated and funds moved to a new wallet? BitGo says it advised Unocoin to do this.

_Data snapshot: Unocoin update of 14 November 2026; BitGo statement of 13 November 2026; Bitquery figures verified 14 November 2026; Chainstack, Socket and Web3Auth/Keplr items dated 3 to 13 November 2026; checked 16 November 2026\. Loss figures are estimates or partial traces and are not added together. Root cause is not established. This is not security or investment advice for any specific setup._
