Fraud Proofs

Cryptographic evidence of operator misbehavior — computable by any party from public data.

A fraud proof is a compact artifact that demonstrates that the operator published an incorrect state root for a specific batch. It contains:

• The pre-batch state snapshot (the state root before the batch)
• The batch transaction log (all requests processed, in order)
• The operator-published post-batch state root
• The locally-computed post-batch state root
• Evidence of divergence (specific accounts or keys where the states differ)

Fraud proofs can detect any of the following operator misbehaviors:

• Incorrect fill execution: The operator matched orders incorrectly — wrong price, wrong size, wrong counterparty.
• Fee theft: The operator charged a higher fee than the configured rate.
• Balance inflation: The operator credited balances without a corresponding deposit transaction.
• Nonce manipulation: The operator accepted orders with duplicate nonces.
• Order book manipulation: The operator applied a different matching algorithm (e.g. not price-time priority).

The fraud proof verifier is included in the Vela repository as the zkvm crate. To verify a specific batch:

# Fetch batch data from DA layer and run verifier
cargo run --bin zkvm-verify -- \
  --batch-id 12345 \
  --da-backend local \
  --da-path ./data/batches.log

In the public beta, fraud proofs are purely informational. The verifier can detect and prove operator misbehavior, but there is no on-chain mechanism to act on the proof. On-chain fraud proof verification, combined with operator slashing, is the M6 milestone.