Reproduce “When AI Takes the Wheel: Security Analysis of Framework-Constrained Program Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yue Liu, Zhenchang Xing, Shidong Pan, Chakkrit Tantithamthavorn
The authors evaluate security in complete Chrome extensions generated by nine models from 140 functional prompts. They use CoCo static analysis to detect potentially dangerous data flows across components with different privileges. DeepSeek-R1’s rate was 49.3% of scenarios with at least one detection, compared with GPT-4o’s 29.3%; these are scenario-level rates, not shares of all generated outputs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.