Surviving in Dark Forest: Towards Evading the Attacks from Front-Running Bots
summary
EVScope uses dynamic binary analysis plus machine learning to scan 6,761,186 arbitrage transactions and 71 real attack transactions and identifies 32 concrete evasion patterns (25 novel), grouped into execution split, access control, profit control and code obfuscation. Example: a 9308/93AE transaction pair on BSC splits state so bots see a revert and omit the attack. Detection relies on trace slicing, control/data-flow checks, and opcode clustering; EVScope beats a decompiler-baseline on F1 and exposes storage-vs-constant tradeoffs for defenders.
AI-written summary · about
chapters (12)
- 0:00Front-running bot attacks
- 1:16Economic impact
- 3:04Challenges overview
- 6:33Evading strategies
- 8:42Access and profit control
- 11:40Unknown evading strategies
- 13:13Evasion techniques findings
- 15:18Execution split case study
- 16:41Code obfuscation
- 17:43Defense comparison
- 19:10Developer recommendations
- 20:13Conclusion
related talks
20mA cat-and-mouse game: how to frontrun a transaction in the future
30mEvasive Maneuvers: Strategies to Overcome Runtime Detection Tools
52mFlying Under the Radar - New Security Evasion Techniques
52mBreaking Fraud and Bot Detection Solutions
41mAPTs Way: Evading Your EBNIDS
12mSpeed, scale & stealth: Securing against ATO events
33mThe Phantoms of the Fraudpera: An Overview of Anti-Detection Tooling
45mAutomatic Discovery of Evasion Vulnerabilities Using Targeted Protocol Fuzzing