Preprint · 2025

CVE-Bench: A Benchmark for AI Agents' Ability to Exploit Real-World Web Application Vulnerabilities

Yuxuan Zhu, Antony Kellermann, Dylan Bowman, Philip Li, Akul Gupta, Adarsh Danda, Richard Fang, Conner Jensen, Eric Ihli, Jason Benn, Jet Geronimo, Avi Dhir, Sudhit Rao, Kaicheng Yu, Twm Stone, Daniel Kang

Abstract

Large language model (LLM) agents are increasingly capable of autonomously conducting cyberattacks, posing significant threats to existing applications. This growing risk highlights the urgent need for a real-world benchmark to evaluate the ability of LLM agents to exploit web application vulnerabilities. However, existing benchmarks fall short as they are limited to abstracted Capture the Flag competitions or lack comprehensive coverage. Building a benchmark for real-world vulnerabilities involves both specialized expertise to reproduce exploits and a systematic approach to evaluating unpredictable threats. To address this challenge, we introduce CVE-Bench, a real-world cybersecurity benchmark based on critical-severity Common Vulnerabilities and Exposures. In CVE-Bench, we design a sandbox framework that enables LLM agents to exploit vulnerable web applications in scenarios that mimic real-world conditions, while also providing effective evaluation of their exploits. Our evaluation shows that the state-of-the-art agent framework can resolve up to 13% of vulnerabilities.

BibTeX Copy

@misc{zhu_cve-bench_2025,
	title = {{CVE}-{Bench}: {A} {Benchmark} for {AI} {Agents}' {Ability} to {Exploit} {Real}-{World} {Web} {Application} {Vulnerabilities}},
	shorttitle = {{CVE}-{Bench}},
	url = {http://arxiv.org/abs/2503.17332},
	doi = {10.48550/arXiv.2503.17332},
	abstract = {Large language model (LLM) agents are increasingly capable of autonomously conducting cyberattacks, posing significant threats to existing applications. This growing risk highlights the urgent need for a real-world benchmark to evaluate the ability of LLM agents to exploit web application vulnerabilities. However, existing benchmarks fall short as they are limited to abstracted Capture the Flag competitions or lack comprehensive coverage. Building a benchmark for real-world vulnerabilities involves both specialized expertise to reproduce exploits and a systematic approach to evaluating unpredictable threats. To address this challenge, we introduce CVE-Bench, a real-world cybersecurity benchmark based on critical-severity Common Vulnerabilities and Exposures. In CVE-Bench, we design a sandbox framework that enables LLM agents to exploit vulnerable web applications in scenarios that mimic real-world conditions, while also providing effective evaluation of their exploits. Our evaluation shows that the state-of-the-art agent framework can resolve up to 13\% of vulnerabilities.},
	urldate = {2026-11-06},
	publisher = {arXiv},
	author = {Zhu, Yuxuan and Kellermann, Antony and Bowman, Dylan and Li, Philip and Gupta, Akul and Danda, Adarsh and Fang, Richard and Jensen, Conner and Ihli, Eric and Benn, Jason and Geronimo, Jet and Dhir, Avi and Rao, Sudhit and Yu, Kaicheng and Stone, Twm and Kang, Daniel},
	month = jun,
	year = {2025},
	note = {arXiv:2503.17332 [cs.CR]},
	keywords = {Computer Science - Artificial Intelligence, Computer Science - Cryptography and Security},
}