The Problem a Key Creates
On a blockchain, possession of a private key is control of the assets at its address. There is no recovery desk at the network level. A single complete key is therefore a single point of failure: whoever obtains it can move the funds, and whoever loses it loses them. Every institutional custody design starts from the requirement that no single person, device, or location can sign alone.
How Threshold Signing Works
Multi-party computation addresses the problem at the level of the key itself. During key generation, each participant produces a share, and the participants jointly derive the public key without any of them learning the others' shares or the full private key. The key exists only in distributed form.
Signing follows the same pattern. A threshold rule, such as two of three or three of five, states how many shares must participate. The participating parties compute with their own shares and exchange intermediate values; the output is an ordinary signature that the network verifies exactly as it would verify one produced from a single key. At no point does the full key appear in memory or on disk anywhere. NIST's Multi-Party Threshold Cryptography project documents the schemes and the security assumptions behind them.
Why Institutions Chose MPC
Two properties make MPC the common choice for institutional custody. First, the signature is standard, so MPC works on any chain, including chains with no native multisignature support, and an MPC-controlled address is indistinguishable on-chain from any other. There is no visible policy to study and no premium in transaction size or fees. Second, the shares can be distributed across people, devices, environments, and locations in whatever configuration policy requires, and they can be refreshed without moving the funds to a new address.
From Cryptography to Controls
MPC decides who can sign; it does not decide whether a transaction should be signed. Institutional programs wrap it in policy. Approval workflows require named roles to initiate and approve. Address allow-lists restrict where funds can go. Per-transaction and daily limits cap what any workflow can move. Monitoring watches for anomalies in timing, size, and destination. Separation of duties ensures that the people who can change policy are not the people who execute under it.
Storage is tiered by exposure. Cold storage holds the bulk of assets with shares on isolated devices and a deliberate procedure to bring them into use. Warm wallets hold the balances needed for the day's settlements, with shares under MPC in separated environments that can sign within minutes under tight limits. Anything hotter is treated as a float, held small and swept often.
Ceremonies, Refresh, and Recovery
Key generation under MPC is a ceremony: a defined procedure, with named participants in separated environments, that produces the shares and the public key without ever producing the private key. The record of that ceremony, including who participated and how each share was stored and backed up, is part of the audit trail.
Shares are not permanent. A refresh procedure re-derives a new set of shares for the same public key, so that a share exposed in the past becomes useless without moving the funds. The same mechanism handles personnel and device changes. Recovery covers the case in which a share or a participant is lost: with a threshold below the total number of shares, signing continues, and the missing share is re-issued through a controlled ceremony. Institutions expect these procedures to be documented in advance and tested on a schedule, not designed during an incident.
Verification
Design is a claim; audit is the evidence. Institutions expect a custody program to be independently audited and penetration tested, and to be able to name the firms and the dates. They ask what share of assets sits in each tier, how limits and allow-lists are changed, what the recovery procedure is if a share or a participant is lost, and how incidents are reported. A provider that answers those questions in writing has turned MPC from a feature into a control.
Frequently Asked Questions
- Is MPC more secure than multisig?
- They solve the same problem, distributed control, in different ways. Multisig keeps whole keys with separate signers and enforces the rule on-chain, which is transparent but chain-dependent and visible. MPC never creates a whole key, produces a standard signature, and works across chains. Security in practice depends more on the operational controls around either mechanism than on the mechanism itself.
- Can an MPC wallet be recovered if a share is lost?
- Threshold schemes are designed so that the loss of fewer shares than the threshold allows does not prevent signing, and shares can be refreshed or re-issued through a controlled ceremony. Institutional programs back up shares in separate secured locations and define the recovery procedure in advance.
- What does warm and cold storage mean under MPC?
- The tiers describe how exposed the key shares are, not the cryptography. Cold storage keeps shares on isolated devices and requires a deliberate procedure to sign. Warm storage keeps shares under MPC in separated environments that can sign within minutes under policy limits, which serves daily settlement.
- How does an institution verify a counterparty's custody program?
- By asking which share of assets sits in each storage tier, how approval workflows and allow-lists are enforced and changed, what monitoring covers the warm tier, who the independent auditors and penetration testers are, and when they last reported. The answers show whether the cryptography is backed by operating controls.
Sources
- Multi-Party Threshold Cryptography — National Institute of Standards and Technology
- Stillman Digital, Custody & Execution Infrastructure — Stillman Digital, Sep 2026