What is a fraud proof?
- Author
- CRYPTO PORT Editorial
- Published
- Updated
- Reading time
- 6 min
In short
A fraud proof is a way to show, after the fact, that a result a layer 2 claimed was wrong. The system optimistically accepts the claim first and then opens a challenge window in which errors can be corrected. That window is exactly why withdrawals from optimistic rollups take time.
Key points
- A mechanism for challenging an incorrect result with evidence
- Claims are accepted first and finalised after a challenge window
- That waiting period is what delays withdrawals
- Its safety assumes someone is actually watching and willing to challenge
Definition
Evidence that a state transition submitted by a layer 2 was incorrect. The submitted result is accepted as valid at first, and verification happens only if a challenge arrives within a set window.
On an optimistic rollup, the operator asserts to layer 1 that processing a batch of transactions produced a particular state. Layer 1 does not verify the claim at that moment. It accepts it on the assumption that someone will object if it is wrong — hence 'optimistic'.
When a challenge is filed, the two sides enter an interactive procedure that repeatedly halves the disputed range of the computation. Eventually a single instruction where they disagree is executed on layer 1 to decide who is right. Verifying one step rather than the whole batch is what keeps the layer 1 cost small.
For any of this to work, there must be participants who independently recompute the chain, and who will post a bond and file a challenge when they find an error. If nobody is watching, a wrong state simply finalises. Which parties are permitted to challenge — a whitelist or anyone — differs from rollup to rollup.
Watch out for
- · Withdrawals through the canonical bridge do not complete until the challenge window closes
- · Where only whitelisted parties may challenge, an error stands if none of them acts
- · Filing a challenge normally requires a bond, which is forfeited if the challenge is wrong