Linux privilege escalation risks detailed with large scale test platform
PrivEscalate: Measuring and Augmenting the Threat of LLM-Automated Linux Privilege Escalation
Cryptography and Security
Summary
Linux privilege escalation is a security problem where an attacker gains more control over a computer than they should. The authors created PrivEscalate, a big set of 531 test situations inside safe containers, to see how well different AI language models can find ways to escalate privileges on Linux systems. They found that the AI's success varies a lot depending on the problem type, the environment, and the way the AI is organized to solve the task. They also built a new helper tool that makes the AI better at this job without changing the AI itself. Their work helps measure risks and can be used to train security teams or test defenses.
What this means in practice
- •For cybersecurity red teams: Use PrivEscalate to simulate Linux privilege escalation attacks and improve realistic training exercises.
- •For security software developers: Employ PrivEscalate to evaluate and strengthen detection or prevention tools against advanced Linux privilege escalation attempts.
Authors
Yixuan Liu, Zilong Zhen, Yin Wu, Yi Li
Abstract
As Large Language Model (LLM) agents increasingly automate offensive operations across the cyber kill chain, their efficacy in complex local post-exploitation tasks remains inadequately quantified. Among these, Linux privilege escalation is a key step between initial access and full system compromise. However, existing evaluations for this task are limited by small sample sizes (fewer than 15 scenarios), lacking the scale to compare model capabilities under executable verification. To address this, we present PrivEscalate, a large-scale benchmark for Linux privilege escalation, comprising 531 Dockerized scenarios spanning 14 sub-categories. We additionally derive 329 parameterized variants to measure sensitivity to environmental distractors. Evaluating six LLMs across three agent architectures reveals: (i) model capability is heterogeneous across vulnerability classes, with no single model dominating across the high-prevalence classes, motivating multi-dimensional risk assessments; (ii) LLM successes are sensitive to environmental perturbation, so configuration rotation can disrupt some exploit attempts but does not eliminate the measured risk; and (iii) agent architectures can materially change success rates and reorder model rankings, though the magnitude is model-dependent. Leveraging these insights, we develop PrivEscAgent, a domain-specialized wrapper that augments a generic ReAct agent with deterministic enumeration, category matching, and step planning. PrivEscAgent improves over prior Linux privilege-escalation agent baselines without underlying LLM modifications. We release PrivEscalate as an open-source, Dockerized measurement instrument supporting LLM agent evaluation, defensive tool validation, and red-team training.