Non-Interactive (NI) ZK Fraud Proofs are pitched as the "best of both worlds" for rollups. However, our latest research finds they risk being the worst of both. We've published in ethresearch an in-depth analysis of NI under adversarial conditions. 🧵
You can read the full, comprehensive analysis on ethresearch: https://ethresear.ch/t/best-of...
Our analysis was motivated by a genuine curiosity to understand if Non-Interactive (NI) fraud proofs could be the path forward for optimistic rollups.
To that end, we "steel-manned" the argument, building the strongest possible case for NI Fraud Proofs before stress-testing it against the established threat model for optimistic rollups (Sybil attacks, censorship, resource exhaustion).
The analysis shows current NI designs don't eliminate the ~7-day withdrawal window of optimistic rollups, and introduce the high capital costs of ZK-rollups. Moreover, defender costs also scale linearly with attacker funds, which is untenable for permissionless systems.
This exercise has reinforced our belief that research in this domain requires a higher standard of rigor. Progress for the entire ecosystem depends on researchers engaging charitably with alternative designs and carefully testing proposals against well-established threat models.
Deep thanks to my coauthor Augusto Teixeira (@AugustoTei17155), reviewers Diego Nehab and @donnoh_eth, and the @cartesiproject team for their support.
Thank you for the kind words ❤️