Glossary

FROST DKG (Distributed Key Generation)

FROST DKG is the setup ceremony where participants jointly generate threshold key shares without a trusted dealer.

Key Takeaways

  • FROST DKG lets a group of participants jointly create threshold signature key shares without any single party ever holding the full private key. No trusted dealer is needed.
  • The protocol runs in two communication rounds: participants broadcast polynomial commitments with Schnorr proofs in round one, then exchange private key shares in round two. After verification, each participant combines received shares into their final signing share.
  • FROST DKG is not fault-tolerant by default: if any participant drops out or sends invalid data, the ceremony must abort and restart. Protocols like Spark use carefully coordinated DKG ceremonies to manage this in practice.

What Is FROST DKG?

FROST DKG (Distributed Key Generation) is the setup phase of the FROST threshold signature scheme. It is the ceremony through which a group of n participants jointly produce a shared group public key and individual signing shares, where any t of the n participants can later cooperate to produce a valid digital signature. The critical property of FROST DKG is that the full private key corresponding to the group public key never exists in any single location: it is generated in a distributed fashion and only ever represented as shares held by individual participants.

FROST DKG builds on Pedersen's classic distributed key generation protocol, enhanced with Feldman's Verifiable Secret Sharing (VSS) and Schnorr proofs of knowledge. These additions prevent rogue-key attacks, where a malicious participant could manipulate the group public key to gain an unfair advantage during signing. The protocol was introduced by Chelsea Komlo and Ian Goldberg in their 2020 paper "FROST: Flexible Round-Optimized Schnorr Threshold Signatures."

While the IETF standardized FROST signing in RFC 9591, the DKG protocol was intentionally left out of scope. The RFC includes only a trusted-dealer key generation method in its Appendix C. This means implementations that want dealerless key generation must follow the original FROST paper or related proposals like PedPop from Blockstream Research.

How It Works

FROST DKG operates in two communication rounds followed by a local computation step. Every participant acts simultaneously as a dealer and a receiver: each one generates a secret polynomial and distributes shares of it to all other participants.

Round 1: Commitment Broadcast

Each participant i generates a random polynomial of degree t-1 (where t is the signing threshold):

// Participant i generates polynomial:
// f_i(x) = a_{i,0} + a_{i,1}*x + a_{i,2}*x^2 + ... + a_{i,t-1}*x^(t-1)
//
// The constant term a_{i,0} is participant i's secret contribution.
// Commitments are elliptic curve points: C_{i,k} = a_{i,k} * G
//
// A Schnorr proof of knowledge of a_{i,0} is also generated
// to prevent rogue-key attacks.

Each participant publishes the following to all other participants over a secure broadcast channel:

  • Elliptic curve commitments to each coefficient of their polynomial (Ci,0, Ci,1, ... Ci,t-1)
  • A Schnorr signature proof of knowledge demonstrating they know the secret value behind their constant-term commitment

The broadcast channel must guarantee that every participant receives the same set of commitments. If participants see different versions of the same commitment, the resulting key shares will be inconsistent. In practice, implementations may use an echo-broadcast sub-protocol to enforce this consistency, which adds an additional communication round.

Round 2: Share Distribution

After collecting and verifying all Round 1 commitments and proofs, each participant evaluates their polynomial at every other participant's index and sends the result as a private share over a secure point-to-point channel:

// Participant i computes a share for participant j:
// share_{i->j} = f_i(j)
//
// This share is sent privately to participant j.
// Only participant j should see this value.

Each participant therefore sends n-1 private shares (one to every other participant) and receives n-1 shares in return.

Verification and Finalization

Upon receiving a share from participant i, participant j verifies it against the commitments that i published in Round 1. This verification uses the homomorphic properties of elliptic curve commitments: the share can be checked without revealing any other participant's secret.

// Participant j verifies the share from participant i:
// share_{i->j} * G  ==  C_{i,0} + j*C_{i,1} + j^2*C_{i,2} + ...
//
// If this check fails, participant i is either
// faulty or malicious, and the protocol must abort.

Once all shares are verified, each participant computes their final signing share by summing all received sub-shares (including their own polynomial evaluated at their own index):

// Participant j's final key share:
// s_j = f_1(j) + f_2(j) + ... + f_n(j)
//
// This is participant j's secret signing share.
// No other participant knows this value.

The group public key is derived by summing the constant-term commitments from all participants:

// Group public key:
// PK = C_{1,0} + C_{2,0} + ... + C_{n,0}
//
// This is the public key that verifiers use to check
// threshold signatures produced by any t participants.

Each participant also computes their verification share (their public signing share), which other participants can use later to verify partial signatures during FROST signing rounds.

Why It Matters

The alternative to DKG is a trusted dealer: a single entity that generates the full private key, splits it into shares, and distributes them. This approach has an obvious weakness: the dealer sees the entire private key and must be trusted to delete it afterward. For high-value applications like multi-party computation wallets and protocol key management, this single point of trust is unacceptable.

FROST DKG eliminates this trust requirement. Because each participant contributes randomness independently, no single party can influence the outcome or reconstruct the full key. This property is essential for protocols like Spark, where operators and users coordinate DKG ceremonies to establish threshold signing keys for virtual UTXOs without any party gaining unilateral spending authority.

FROST DKG's two-round structure also makes it more practical than older DKG protocols that required three or more rounds. Fewer rounds mean lower latency and simpler coordination, which matters when DKG ceremonies run frequently at scale.

Use Cases

Threshold Wallets

MPC wallets and threshold custody solutions use FROST DKG to set up signing keys across multiple devices or parties. A 2-of-3 wallet, for example, runs DKG among three participants so that any two can sign transactions. The wallet provider never possesses the full key, providing genuine self-custody guarantees.

Protocol Key Management

Layer 2 protocols and sidechains use FROST DKG to establish operator signing committees. When a new set of operators rotates in, a fresh DKG ceremony generates new key shares without exposing the underlying signing key. This is how systems like Spark establish cooperative signing authority for managing user deposits.

Bridge Security

Cross-chain bridges that hold funds in threshold-controlled addresses use DKG to distribute custody across multiple independent validators. This prevents any single validator from stealing bridged assets. FROST DKG is preferred over trusted-dealer setups here because the stakes are high and no single entity should ever see the full bridge key.

Signing Committees

Organizations that require multiple approvers for sensitive operations (treasury management, smart contract upgrades, certificate issuance) can use FROST DKG to establish a threshold signature scheme. The resulting signatures are indistinguishable from single-signer Schnorr signatures on the blockchain, providing both security and privacy.

Risks and Considerations

No Built-in Fault Tolerance

Standard FROST DKG is not robust: if any participant drops out, sends an invalid share, or fails to broadcast their Round 1 commitment, the entire ceremony must abort and restart. This contrasts with more complex DKG protocols (such as the Gennaro-Jarecki-Krawczyk-Rabin variant) that can tolerate up to n-t faulty participants by adding a complaint round for identifying and disqualifying bad actors.

In practice, this means FROST DKG works best when participants are reliable and have stable network connectivity. For large participant sets or adversarial environments, implementations may add application-level retry logic or complaint mechanisms on top of the base protocol.

Broadcast Channel Assumption

Round 1 requires a secure broadcast channel: a communication mechanism that guarantees every participant receives exactly the same message. If an attacker can show different commitments to different participants (an equivocation attack), the resulting key shares will be inconsistent, potentially allowing the attacker to forge signatures.

Implementing a true broadcast channel is non-trivial. Solutions include echo-broadcast protocols (where participants relay what they received and compare), consensus mechanisms, or simply relying on a trusted coordinator to relay messages honestly. Each approach introduces different trust assumptions and latency tradeoffs.

Rogue-Key Attacks

Without the Schnorr proof of knowledge in Round 1, a malicious participant could choose their commitment as a function of other participants' commitments, manipulating the group public key to a value they can sign for alone. The proof of knowledge prevents this by requiring each participant to demonstrate they actually know the secret behind their commitment before seeing others' values.

This is why the order of operations matters: commitments and proofs must be broadcast simultaneously, or at minimum, commitments must be fixed (via a commit-reveal scheme) before proofs are verified. Implementations that allow participants to adaptively choose their commitments after seeing others are vulnerable to this attack.

Secure Channel Requirements

Round 2 shares must be sent over encrypted, authenticated point-to-point channels. If an attacker intercepts these shares, they learn partial information about the final key shares. If they intercept enough shares from different participants, they may be able to reconstruct a participant's full signing share.

Key Share Lifecycle

After DKG completes, participants must securely store their key shares and the associated verification data. If a participant loses their share, it cannot be recovered without running a new DKG ceremony or using a proactive secret sharing protocol to refresh shares. There is no "forgot my password" path: the distributed nature of the key means no one can regenerate a lost share.

This glossary entry is for informational purposes only and does not constitute financial or investment advice. Always do your own research before using any protocol or technology.