When an Intelligent Sequencer Guards a Chain
summary
Zircuit deploys an EVM-equivalent ZK rollup that runs AI-powered sequencer-level security to detect and quarantine exploit transactions before they make chain state, claiming models miss one in 200 hacks and protecting a network handling roughly $1–3.6B. The architecture mixes SLS 1.0 transaction-level pre-block scanning with block-aware analysis and an optimistic SLS 3.0 reorg path, uses migrated real-world hacks and a vulnerable DeFi corpus for testing, and addresses sync, forced-inclusion, DOS and false-positive handling.
AI-written summary · about
chapters (40)
- 0:00Introduction to Martin Derka and Zircuit
- 0:47The problem: Recurring blockchain hacks
- 1:25Zircuit's mission: Proactive chain security
- 1:57Zircuit as EVM-equivalent ZK rollup
- 2:34Traditional rollup flow vs SLS architecture
- 3:06How sequencer-level security works
- 3:37Transaction quarantine process
- 4:08System design requirements
- 4:40Scanning transactions vs full blocks
- 5:10SLS 1.0: Transaction-level analysis
- 5:42Advantages of pre-block analysis
- 6:16Transaction dependency challenges
- 6:43SLS 2.0: Block-level analysis
- 7:11Performance tradeoffs
- 7:32SLS 3.0: Optimistic execution
- 8:04Handling analysis delays
- 8:28Zircuit's hybrid SLS approach
- 8:48Current architecture evolution
- 9:24Training AI to detect hacks
- 10:01Leveraging EVM compatibility
- 10:33Model accuracy metrics
- 11:13Potential impact on historical hacks
- 11:39False positive challenges
- 12:09Testing methodology
- 12:43Porting real-world hacks for testing
- 13:14Migration challenges
- 13:44Research on contract migration
- 14:17Address reference handling
- 14:49Phishing protection use case
- 15:19Wallet security limitations
- 15:50On-chain phishing detection
- 16:22Superiority of sequencer-level protection
- 16:54Rule-based transaction filtering
- 17:25Institutional adoption benefits
- 17:58KYC/AML compliance solutions
- 18:30System synchronization challenges
- 18:59Forced inclusion considerations
- 19:32DOS attack prevention
- 20:03User experience design
- 20:36Future detector development
related talks
07mEnhancing L2 Security with Sequencer-Level Protection
20mUsing AI to Secure Dapps and Users
23mBuilding On Zircuit, A Hackathon Introduction
21mBuilding on Zircuit A Hackathon Introduction
16mExploring Attacks in the Modular Rollup World
22mA Multi-Layer Dynamic Security Framework for DeFi Smart Contracts
25mWho Will Fight For ETH
20mA cat-and-mouse game: how to frontrun a transaction in the future