Reproduce “PrivEscalate: Measuring and Augmenting the Threat of LLM-Automated Linux Privilege Escalation” 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
Yixuan Liu, Zilong Zhen, Yin Wu, Yi Li
PrivEscalate measures Linux privilege-escalation agents on 531 Dockerized scenarios. Six models start from unprivileged SSH footholds, and success means reaching root within 20 steps. Results vary with agent framework and environment: PrivEscAgent raises Claude Sonnet 4.6’s success rate from 37.7% to 71.9%. Because scenarios are weighted toward Sudo and SUID/SGID, these scores do not predict real-world compromise rates.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.