arXiv cs.CL· Terry Yue Zhuo, Dingmin Wang, Hantian Ding, Varun Kumar, Zijian Wang·· 2 天前AI 评分45
Cyber-Zero:无需运行时环境训练网络安全智能体
Cyber-Zero: Training Cybersecurity Agents without Runtime
AI 导读
Cyber-Zero 是首个无需运行时环境的框架,利用公开 CTF writeups 和人格驱动 LLM 模拟逆向还原运行时行为,合成高质量智能体轨迹用于训练网络安全 LLM。
正文 · 原文
Abstract:Large Language Models (LLMs) have achieved remarkable success in software engineering tasks when trained with executable runtime environments, particularly in resolving GitHub issues. However, such runtime environments are often unavailable in other domains, especially cybersecurity, where challenge configurations and execution contexts are ephemeral or restricted. We present Cyber-Zero, the first runtime-free framework for synthesizing high-quality agent trajectories to train cybersecurity LLMs. Cyber-Zero leverages publicly available CTF writeups and employs persona-driven LLM simulation to reverse-engineer runtime behaviors and generate realistic, long-horizon interaction sequences without actual environments. Using trajectories synthesized by Cyber-Zero, we train LLM-based agents that achieve up to 11.4% absolute performance gains over baseline models on three prominent CTF benchmarks: InterCode-CTF, NYU CTF Bench, and Cybench. Our best model, Cyber-Zero-32B, establishes new state-of-the-art performance among open-weight models, matching the capabilities of proprietary systems like DeepSeek-V3-0324 and Claude-3.5-Sonnet while offering superior cost-effectiveness, and demonstrating that runtime-free trajectory synthesis can effectively democratize the development of state-of-the-art cybersecurity agents.
| Comments: | Public Link: this https URL |
| Subjects: | Cryptography and Security (cs.CR); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2508.00910 [cs.CR] |
| (or arXiv:2508.00910v3 [cs.CR] for this version) | |
| https://doi.org/10.48550/arXiv.2508.00910 arXiv-issued DOI via DataCite |
Submission history
From: Terry Yue Zhuo [view email]
[v1]
Tue, 29 Jul 2025 18:10:37 UTC (583 KB)
[v2]
Mon, 25 Aug 2025 18:12:25 UTC (590 KB)
[v3]
Fri, 13 Nov 2026 11:42:17 UTC (583 KB)
来源:arXiv cs.CL · arxiv.org