Get Started
Home
Topics
Search
Library
Research questionHow can we build diverse, validated ground-truth datasets for smart-contract vulnerability detectors without manual construction?Manually creating vulnerable contracts is expensive, while naïve modifications may fail to compile, alter intended behavior, or fail to contain the target vulnerability. Useful datasets therefore need validated examples that preserve contract semantics while covering varied vulnerability types and structures.
AI
Code Generation & Program Synthesis
Evaluation & Benchmarks
Research Paper
Technology
Latest papersRecent research connected to this question, newest first.Automated Vulnerability Injection in Smart Contracts Using Large Language ModelsThe evidence concerns LLM-based vulnerability injection into real-world Solidity contracts, targeting 49 OpenSCV vulnerability types. Candidates were checked for compilation, execution, business logic preservation, and the intended vulnerability; from nearly 1,000 generated variants, 32 validated contracts spanning 25 types remained. The results also show limitations from non-determinism, semantic preservation, scalability, and diversity, and were used to assess three static analyzers.research paper · Sep 2, 2026
Related questions
How should ML vulnerability-detection benchmarks measure practical security capabilities beyond narrow binary function-level tasks?How can vulnerability-fixing history be reused to detect and repair similar undisclosed flaws in other code?How can PII detectors avoid missed entities when deployment data shifts from benchmark conditions?How can we select nonredundant preference data across safety datasets without losing robustness?