Wallet custody · A visual comparison
Who can sign?
Who can recover?
A wallet’s security is more than where its key lives. Compare four ways to organise signing, then test what happens when a device is compromised, someone is unavailable or an approval goes wrong.
Read these as four designs, not a universal safety ranking. More separation can reduce one exposure while adding recovery work and delays.
Illustrative · fictional signing arrangements
Choose a design, then change what happens. Keep the same situation selected to compare the trade-offs.
Signing threshold: 1 of 1 · no offline signer required in this example
Signing rule · when the threshold is met
1 required signer → one transaction signatureThe wallet signs on the everyday device, then sends the signed transaction.
Result under these assumptions
The intended transaction can be signed.
All required participants are available and approve the intended transaction. This configuration requires 1 of 1 signing participants.
What it adds
Simple access with few signing steps.
What it asks of you
The device and a complete recovery backup are each routes to signing control.
This model does not assess a product, generate keys, perform MPC or move assets. It assumes correct cryptographic implementation and the specific compromise described. Backup design, recovery rights, firmware, physical access, independent approvals and transaction policy remain separate questions.
Four designs, four sets of decisions
1 · A software wallet on your everyday device
The device holds usable signing material. A mnemonic is a recovery representation of a secret, not a requirement that it be kept on the same device. Keeping the only backup there also couples device loss with loss of recovery.
2 · A separate hardware signer
In normal signing, the connected computer supplies transaction data and the device returns a signature. Public data can also travel; “only a signature leaves” is too broad. Bluetooth, USB or optical transfer describes transport. Key isolation, trustworthy transaction display and recovery handling determine more than the transport label alone.
3 · Distributed signing with MPC
A threshold-signing protocol lets enough participants sign using shares. With suitable distributed key generation and signing, no participant needs to hold the full key. That property must also be checked against how keys are imported, backed up and recovered. MPC is not simply splitting a seed and later joining it again.
4 · Make an offline participant necessary
Keeping a share on an air-gapped signer changes its network exposure. The threshold and policy must require that signer if every transaction is meant to pass through it. Offline signers still receive data, run software and can approve the wrong transaction. Additional separation costs time and can reduce availability.
Try a different arrangement
Why can a design improve resistance to one compromised device yet make recovery or day-to-day signing harder?
Your note stays in this page’s memory. It is not sent, saved or graded.
Before choosing a design
Ask who can authorise a transfer, which devices must participate, what happens when one person is unavailable, and who can use the recovery route. Technical signing control and legal entitlement are separate; succession requires both to be addressed.
Start with the threats and operating needs. Then test the signing and recovery procedures, including the things that could go wrong.
Read the sources
- NIST: multi-party threshold cryptography: distributing trust in cryptographic operations.
- Ledger: Nano X Bluetooth security model: a vendor-specific explanation of transport and key isolation, not a claim about all hardware wallets.
- Fireblocks: air-gapped MPC cold wallet: one commercial implementation; not the fictional architecture illustrated above.
Working through these decisions for a team? How I approach custody