Threshold Signatures and MPC
How t-of-n threshold signatures and multi-party computation let multiple parties sign together without any one holding the full key — the cryptography behind Fireblocks and modern institutional custody.
The basics
The Problem With Single Keys
A private key controlling significant value is a single point of failure. If it leaks, all funds are gone. If it's lost, all funds are stuck. Multi-sig wallets solve part of this: instead of one key, require N keys to sign, of which at least M must agree. But traditional multi-sig (like Bitcoin's CHECKMULTISIG or Ethereum's Safe) reveals on-chain which N parties signed — privacy-leaking and gas-expensive. Threshold signatures fix both issues.
What Threshold Signatures Do
A t-of-n threshold signature scheme distributes pieces of a single private key across N participants such that any T of them can together produce a valid signature, but fewer than T cannot. From the outside, the resulting signature looks exactly like a single-key signature — same format, same on-chain footprint. The fact that N parties were involved is invisible to anyone observing the chain.
Why It Matters in Practice
Institutional crypto custodians such as Fireblocks and BitGo use MPC threshold signing — your funds are protected by key shares held by separate parties and systems, but the chain just sees a single ECDSA signature. Consumer wallets like ZenGo use MPC signing too, and a 2023 Fireblocks research paper lists Coinbase and Binance among the platforms whose wallets are powered by MPC signing protocols.
- Single key: easy but a single point of failure
- Traditional multi-sig: N-of-M but visible on-chain, expensive
- Threshold MPC: N-of-M but indistinguishable from single-key on-chain
- Used by: institutional custody, Fireblocks, ZenGo, BitGo, large exchange wallets
Key Takeaways
- Threshold signatures split one private key into shares; any T-of-N shares can co-sign
- From the chain's perspective, the output is a normal single-key signature
- Used heavily by institutional custody and modern multi-party wallet products
- The privacy and gas advantages over on-chain multi-sig are significant
Going deeper
Distributed Key Generation (DKG)
The first hard problem is generating a key that no single party ever held. Distributed Key Generation protocols let N participants collaboratively compute key shares such that the full key is never assembled anywhere — it only exists implicitly. Each participant's share, combined with T-1 others, can sign; but no single participant can extract the full key. This eliminates a class of operational vulnerability where the 'key generation ceremony' itself is the weakest link.
Signing Protocols
Once keys are distributed, threshold signing requires the participating shareholders to run a multi-round protocol to produce a signature. Each scheme is curve-specific: GG18/GG20 (Gennaro-Goldfeder) for ECDSA, FROST for Schnorr (Bitcoin Taproot), and various BLS-specific protocols. The protocols differ in round count, communication cost, and which dishonest behaviours they tolerate. Choosing a scheme is non-trivial — production deployments lean on well-vetted libraries (e.g., Coinbase's open-source cb-mpc library or Fireblocks' open-source mpc-lib).
Multi-Party Computation Generally
MPC threshold signatures are one application of a broader cryptographic primitive: secure multi-party computation. MPC lets N parties jointly compute any function on inputs they keep private — auctions where bids stay secret, training ML models without exposing training data, etc. Threshold signing is a small, well-understood subset. Other MPC use cases (private DeFi, sealed-bid auctions) are emerging but less mature.
Threshold vs Smart-Contract Multi-sig
ERC-4337 smart-contract wallets and Safe Multisig let multiple EOAs co-sign. This is on-chain multi-sig — every signer is visible. Threshold MPC happens off-chain and produces a single on-chain signature. They serve different use cases: smart-contract multi-sig is transparent and programmable (you can add timelocks, spending limits); threshold MPC is private and gas-efficient. Many sophisticated setups use both layers.
Key Takeaways
- DKG generates a private key that no single party ever holds — eliminates the key-generation choke point
- Threshold signing is curve-specific: GG18/GG20 for ECDSA, FROST for Schnorr, various for BLS
- Threshold signatures are one application of broader MPC; other MPC use cases (private auctions, sealed bidding) are emerging
- Threshold MPC and smart-contract multi-sig are complementary — many setups use both
Under the hood
Adversary Models — Honest, Semi-Honest, Malicious
MPC security is precisely specified by an adversary model. Semi-honest (or honest-but-curious): parties follow the protocol but try to learn private inputs from messages. Malicious: parties can deviate arbitrarily — send malformed messages, abort, collude. Protocols proving security against malicious adversaries are dramatically more expensive than semi-honest variants. Production threshold signing schemes target malicious adversary models with abort detection — if a party deviates, the protocol detects it and can identify the culprit.
Proactive Refresh and Share Rotation
A long-lived threshold key is vulnerable to slow compromise — an attacker who steals one share per year over T years eventually gains enough to sign. Proactive refresh protocols periodically rerandomise the shares without changing the underlying public key: each share looks different after refresh, but the combined T-of-N still signs the same key. Stolen old shares become useless. Production systems can run refresh on a schedule as part of operational hygiene.
Threshold ECDSA Is Surprisingly Hard
ECDSA's signing equation involves a multiplicative inverse modulo the curve order — easy with a single key, very hard to compute in a threshold setting. Practical threshold ECDSA only became feasible around 2017-2018 with GG18 (Gennaro-Goldfeder) and similar schemes. The protocols have non-trivial round counts, often rely on Paillier homomorphic encryption (which is slow, though the DKLs protocols avoid it), and have had several discovered vulnerabilities requiring fixes. By contrast, threshold Schnorr (FROST) is dramatically simpler — two signing rounds, much faster. This is part of why Bitcoin's Taproot upgrade matters for threshold setups.
Verifiable Secret Sharing Foundations
The underlying primitive is Shamir's secret sharing (1979): a polynomial of degree T-1 over a finite field has T points that uniquely determine it; any subset of T evaluations reconstructs the polynomial, but T-1 reveal nothing. Threshold MPC schemes extend this with verifiable secret sharing (VSS) so participants can confirm they received valid shares without seeing the secret, and with publicly verifiable shares so observers can audit the process. This stack of primitives is what makes threshold cryptography practical.
MPC Library Quality Varies
Threshold ECDSA libraries have a documented history of subtle vulnerabilities — a 2023 paper (ACM CCS 2024) presented four practical key-extraction attacks on popular threshold-ECDSA protocols and implementations, and the authors of GG20 have stated that it is obsolete and should not be used. The math is unforgiving and the protocols are complex. Treat MPC implementations as you would cryptographic primitives generally: use well-audited libraries from reputable vendors, not rolled-from-scratch implementations. The cost of getting it wrong is catastrophic.
Key Takeaways
- Malicious-adversary security with abort detection is the production standard
- Proactive refresh rotates shares periodically — defeats slow compromise
- Threshold ECDSA is much harder than threshold Schnorr — Taproot enables much simpler setups
- Verifiable secret sharing is the foundational primitive — Shamir 1979 still load-bearing
- Use audited libraries; threshold MPC has a history of subtle implementation bugs
Related Content
More Topics
Browse all 189 topicsFinality and Fork Handling
What it means for a transaction to be 'final,' why different chains finalise differently, how reorgs happen, and the MEV incentives that complicate the picture.
State Pruning and Expiry
Why Ethereum's node storage needs have grown to terabytes, what 'archive,' 'full,' and 'light' nodes actually store, and how proposals like EIP-4444 and new state-tree designs (verkle or binary) aim to keep node hardware accessible for years to come.
ECDSA vs EdDSA vs BLS
The three signature schemes you'll meet across the crypto stack — what each does well, what tradeoffs they impose, and why Ethereum uses three of them simultaneously.
zk-SNARK vs zk-STARK vs PLONK
How the three major proof-system families compare on trusted setup, proof size, prover cost, and quantum resistance — and which production rollups picked which.
Verifiable Random Functions
How VRFs produce randomness that's both unpredictable before commitment and cryptographically verifiable after — enabling fair lotteries, leader election, and on-chain randomness without trusted parties.
Hash Functions Compared
SHA-256, Keccak-256, Blake3, and Poseidon — which one each chain uses, why ZK systems needed a new family of 'arithmetic-friendly' hashes, and what tradeoffs each makes.
Stealth Addresses and Confidential Transactions
Privacy primitives that hide who's receiving what — from Monero's foundational stealth addresses to Ethereum's ERC-5564 and the legal context post-Tornado-Cash.
CDP Lifecycle
How collateralized debt positions work end-to-end — minting DAI against ETH, paying stability fees, surviving liquidations, and the structural lessons from MakerDAO's Black Thursday and Liquity's no-fee model.
References & further reading
- secondary
- secondary
- primaryIETF - RFC 9591: The Flexible Round-Optimized Schnorr Threshold (FROST) Protocol for Two-Round Schnorr Signatures
Standardised threshold Schnorr signing.
- primaryMakriyannis, Yomtov, Galansky (Fireblocks) - Practical Key-Extraction Attacks in Leading MPC Wallets (ePrint 2023/1234, ACM CCS 2024)
Four key-extraction attacks on threshold-ECDSA protocols and implementations.