fraud proof

A fraud proof is the mechanism that lets anyone prove to the base chain that an optimistic rollup published an incorrect state transition, so the bad result can be rejected. It flips the usual burden of proof: instead of requiring the rollup to prove every batch is correct (the ZK approach), the system assumes correctness and only acts when a challenger demonstrates a specific error. This is what makes 'optimistic' security real rather than mere trust — one honest watcher with the published data can overturn a fraudulent claim.

Naively, a fraud proof would re-execute the whole disputed batch on Ethereum, but that is far too expensive — defeating the point of moving computation off-chain. The dominant solution is the interactive bisection (or dispute) game. The challenger and the defender are forced to repeatedly halve the disputed execution trace: 'we agree on the state at step N, disagree at step 2N — so what is the state at step 1.5N?' Each round narrows the disagreement until it isolates a single instruction. Ethereum then executes only that one step in a minimal verifier, compares the result, and rules. Arbitrum pioneered multi-round interactive proofs; Optimism's Cannon and the OP Stack fault-proof system run a MIPS-emulated single-step check.

Three conditions make a fraud proof effective. The transaction data must be available so the challenger can compute the correct trace at all; the challenge period must be long enough to detect and submit before withdrawals finalize; and the right to challenge must be permissionless, since restricting it to a whitelist reintroduces trust. A subtle attack is therefore to spam or censor the challenger out of the window, or to withhold data so no honest proof can be built — which is why data availability and censorship resistance are inseparable from fraud-proof security.