WingFuzz: Implementing Continuous Fuzzing for DBMSs
summary
WingFuzz implements continuous DBMS fuzzing with a grammar-derived mutator, rapid-commit fuzzing that targets only functions affected by commits, and a noise-resilient anomaly assessment that isolates threads, dumps stacks, wipes databases and restarts the server to avoid cross-test noise. It finds 236 previously unknown bugs across 12 enterprise DBMSs (232 confirmed), outperforms three state-of-the-art fuzzers with 41–85% more branch coverage, and adds ~400k branches and +5 bugs from grammar mutations; ClickHouse CI adopts it.
AI-written summary · about
chapters (9)
- 0:00WingFuzz introduction
- 1:50Continuous fuzzing steps
- 3:17Fuzzing DBMS obstacles
- 5:25WingFuzz design
- 7:10Evaluating on DBMSs
- 8:57Component effectiveness
- 10:21Vendor feedback
- 10:43Discovered vulnerabilities
- 11:46Conclusion
related talks
31m“Fuzz or lose..."
01h 15mMake your programs more reliable with Fuzzing
39mFuzzing Session: Finding Bugs and Vulnerabilities Automatically
39mMaking Fuzzing Part of Your Software Development Lifecycle
11mIDFuzz: Intelligent Directed Grey-box Fuzzing
53mFuzz Testing for Fun and Profit
01h 00mClusterFuzz: Fuzzing at Google Scale
53mFuzzing 101