Learning to Boost Disjunctive Static Bug-Finders

Explore a cutting-edge learning-based approach for enhancing disjunctive static bug-finders in this 41-minute conference talk from ACM SIGPLAN. Delve into the challenges of path-sensitive analysis in industrial static bug-finders and discover how machine learning techniques can be leveraged to develop efficient state-selection heuristics. Learn about innovative strategies for collecting alarm-triggering traces, training multiple candidate models, and adaptively selecting the most appropriate model for each target program. Gain insights into improving the cost-to-efficiency ratio of modern static bug-finders and minimizing false positives in bug reports.

[INFER'23] Learning to Boost Disjunctive Static Bug-Finders