Glossary

Abort Identification

Abort identification detects which participant caused a multi-party signing session to fail, enabling accountability and forward progress in threshold signature protocols.

Key Takeaways

  • Abort identification (also called identifiable abort) is a security property of threshold signature protocols that pinpoints which participant caused a signing session to fail.
  • Without abort identification, a single malicious signer can indefinitely block a group from producing valid signatures, because honest participants cannot determine who to exclude from the next attempt.
  • FROST supports identifiable aborts by allowing the coordinator to verify each signing share individually, while protocols like ROAST build on this property to guarantee that signing always completes.

What Is Abort Identification?

Abort identification is a security property of multi-party cryptographic protocols. When a signing session fails because one or more participants submitted invalid data, abort identification means the protocol can determine exactly which participant misbehaved. The concept originates in secure multi-party computation (MPC) research and was formalized by Ishai, Ostrovsky, and Zikas in 2014, who required that an abort reveal at least one malicious party's identity.

In the context of threshold signatures, abort identification matters because producing a valid group signature requires cooperation among multiple signers. If any signer submits a corrupted partial signature, the aggregate signature will be invalid. A protocol with abort identification lets honest participants identify the bad actor, exclude them, and retry with a new set of signers. Without this property, the group has no way to make progress: a single adversary can disrupt every signing attempt without consequence.

How It Works

The mechanism behind abort identification varies by protocol, but the core idea is consistent: each participant's contribution must be individually verifiable. If the final result is invalid, the coordinator (or participants) can check each contribution in isolation to find the source of failure.

Verification of Partial Signatures

In a Schnorr-based threshold signing protocol like FROST, each signer produces a partial signature (a signing share) during Round Two. The coordinator can verify each share against the signer's public key share and commitment. If a share fails verification, the coordinator knows exactly which signer produced it.

As described in RFC 9591 (the FROST specification), the protocol includes a dedicated identifiable abort procedure in Section 5.4. The coordinator aggregates all signature shares and, if the resulting aggregate signature is invalid, checks each individual share to find the culprit.

In practice, an optimized coordinator can first verify the final aggregate signature. If it passes, all shares were valid and no per-share verification is needed. The more expensive per-share checks run only when aggregation fails:

  1. The coordinator collects all signing shares from participants in the signing group
  2. The coordinator attempts to aggregate the shares into a complete Schnorr signature
  3. If the aggregate signature verifies against the group public key, signing succeeds
  4. If verification fails, the coordinator runs the identification protocol: it checks each participant's signing share against that participant's public key share and commitment
  5. Any share that fails individual verification identifies a misbehaving signer
  6. The misbehaving signer is excluded, and signing restarts with a replacement participant (if one is available)

Commitments and Binding

Abort identification relies on cryptographic commitments. During Round One of FROST, each signer broadcasts a commitment to their nonce values. These commitments serve two purposes: they prevent signers from adaptively choosing their nonces after seeing others' contributions, and they provide the reference point against which signing shares are later verified.

FROST uses a binding factor derived from the set of all commitments and the message being signed. This ensures that each signer's share is bound to a specific signing session. A signer cannot reuse a commitment from a different session or modify their contribution after seeing other signers' commitments.

Pseudocode Example

The following pseudocode illustrates how a coordinator identifies a misbehaving signer after aggregation fails:

// After collecting all signing shares
aggregate_sig = aggregate(signing_shares)

if verify(aggregate_sig, group_public_key, message):
    return aggregate_sig  // Success: all shares were valid

// Aggregation failed: identify the bad signer
for each (signer_id, share) in signing_shares:
    pub_key_share = get_public_key_share(signer_id)
    commitment = get_commitment(signer_id)

    if not verify_share(share, pub_key_share, commitment, message):
        report_misbehavior(signer_id)
        exclude_from_future_sessions(signer_id)

// Retry signing with remaining honest signers

Why It Matters

Without abort identification, threshold signing protocols are vulnerable to a simple denial-of-service attack: a malicious signer submits garbage data, the aggregate signature fails, and the honest signers have no idea who caused the failure. The group must either retry with the same set (risking the same disruption) or exclude signers at random (potentially excluding honest participants).

This matters especially in real-world deployments where signing must complete reliably. Consider a Bitcoin custody solution using a 3-of-5 threshold signature scheme. If one key holder is compromised and starts disrupting signing sessions, the other participants need to identify and remove that signer to continue operating. Abort identification makes this possible without trusting any third party.

Protocols like Spark's FROST-based signing system rely on abort identification to maintain liveness: the ability to always produce a valid signature when enough honest signers are available. By quickly identifying and excluding misbehaving participants, the protocol can retry and succeed without manual intervention.

Use Cases

Threshold Custody

Institutional custody solutions use threshold signatures so that no single party holds a complete private key. Abort identification ensures that a compromised or malfunctioning signing device does not permanently block withdrawals. The coordinator identifies the faulty device, switches to a backup, and completes the signature.

Distributed Key Generation Ceremonies

During a distributed key generation (DKG) ceremony, participants jointly create key shares without any single party seeing the full key. If a participant submits invalid data during the ceremony, abort identification lets the group identify and exclude them, then restart the DKG with a clean set.

Signing Coordinators and ROAST

The ROAST protocol takes abort identification a step further. ROAST runs multiple concurrent FROST signing sessions with different subsets of signers. When one session aborts, the coordinator uses FROST's identifiable abort property to exclude the misbehaving signer from future sessions. Because ROAST maintains parallel sessions, at least one is guaranteed to contain only honest signers and will eventually produce a valid signature.

Layer 2 Protocols

Bitcoin layer 2 protocols that use multi-party signing for state updates benefit from abort identification. In systems like Spark, where operators and users cooperatively sign transactions, abort identification ensures that an unresponsive or malicious operator cannot silently block state transitions. The protocol can detect the failure, attribute it, and take corrective action.

Comparison With Other Approaches

Security With Abort (No Identification)

Many MPC protocols offer only "security with abort": if a malicious party disrupts the protocol, honest parties can detect that something went wrong, but they cannot determine who caused the failure. This is a weaker property. The protocol halts safely (no secrets are leaked), but the group has no mechanism to remove the attacker and retry.

Robust Protocols (Guaranteed Output Delivery)

At the other end of the spectrum, robust protocols guarantee that the output is always produced, regardless of adversary behavior. This is strictly stronger than identifiable abort. However, robust threshold signing protocols typically require more communication rounds, higher computational cost, or stronger trust assumptions. FROST chose identifiable abort over full robustness as a practical trade-off: two rounds of communication and efficient Schnorr arithmetic, with the ability to identify and exclude bad actors when needed.

FROST vs. MuSig2

MuSig2 is a multi-signature (not threshold signature) protocol. In MuSig2, all participants must cooperate. The protocol supports detection of invalid partial signatures, but its design differs because all signers are required (there is no threshold subset). FROST extends this concept to the threshold setting, where any t-of-n subset can sign, and the identification mechanism works across arbitrary subsets of signers.

Risks and Considerations

Coordinator Trust

In many FROST deployments, a coordinator aggregates commitments and shares. The coordinator learns which shares are valid or invalid, giving it the power to falsely blame an honest signer. Mitigations include broadcasting all shares to every participant so they can independently verify, or rotating the coordinator role. The original FROST paper notes that a malicious coordinator can prevent signing from completing but cannot forge signatures or extract secret key material.

Cost of Identification

Per-share verification is more computationally expensive than simply checking the aggregate signature. In the optimized flow (verify aggregate first, check shares only on failure), this cost is paid only when an abort occurs. However, under active attack where aborts happen frequently, the repeated per-share verification becomes a significant overhead.

Equivocation Attacks

A malicious signer might send different commitments to different participants (equivocation). If participants receive inconsistent views of the commitment set, the identification protocol may produce conflicting results. This is why FROST requires a consistent broadcast channel for commitments, or at minimum, agreement on the commitment list before signing proceeds.

Liveness vs. Safety

Abort identification improves liveness (the ability to eventually produce a signature) but does not guarantee it. If the adversary controls enough signers to prevent forming a threshold quorum of honest participants, no amount of identification will help. The protocol remains safe (no forged signatures), but it cannot produce output. ROAST addresses this gap by guaranteeing output delivery as long as at least t honest signers are available.

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.