LLMs like GPT-3.5 and GPT-4 can obscure code authorship, study finds

A new study published on arXiv explores how large language models like GPT-3.5 and GPT-4 can be used to obscure code stylometry, a technique used for authorship attribution and cybersecurity. Researchers found that LLMs can alter code to avoid detection by classifiers, with effectiveness varying based on prompt engineering strategies and whether single-shot or multi-shot methods were employed. The study also assessed the models' ability to preserve code functionality after modifications, highlighting challenges in maintaining code integrity. AI

IMPACT This research highlights potential vulnerabilities in code authorship attribution techniques, impacting cybersecurity and software engineering practices.

RANK_REASON The cluster contains a research paper published on arXiv detailing a comparative study of LLM capabilities. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• GPT-3.5
• GPT-4
• Hugging Face
• large-language models
• Random Forest
• ScienceCast

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →