CTF-Dojo - Training Language Model Agents to Find Vulnerabili

项目说明

Training Software Agents to Find with CTF-Dojo

<p align="left">
<a href=" src="
</p>

<p align="left">
🧐&nbsp;<a href="#o">O</a>
| 🚀&nbsp;<a href="#quick-start">Quick Start</a>
| 📜&nbsp;<a href="#citation">Citation</a>
</p>

CTF-Dojo is a large-scale executable runtime for training LLM agents with feedback. It provides 658 fully functional CTF-style challenges, each containerized with Docker for guaranteed reproducibility, and is automatically assembled —an end-to-end pipeline that converts public artifacts into ready-to-run en in minutes. Training on just 486 high-quality, execution- trajectories yields up to 11.6% absolute gains o baselines across InterCode-CTF, NYU CTF Bench, and Cybench; our best 32B model reaches 31.9% Pass@1, ri frontier systems. These results show execution-grounded signals are pivotal for ad powerful ML agents without relying on proprietary infrastructure.

O

<p align="center">
<img src="asset/tease.svg" alt="CTF-Dojo Teaser" width="800"/>
</p>

Quick Start

1. Clone the Pwn.College's CTF Archive as a template

2. Run CTF-Forge on the CTF Archive to create CTF-Dojo

3. Collect writeups (external dataset)

4. Create the metadata for CTF-Dojo challenges

5. Map writeups to CTF-Dojo challenges

6. Collect trajectories from CTF-Dojo challenges

Citation

If you use this benchmark suite in your research, please cite:

License

This project is licensed under the CC-BY-NC-4.0 - see the LICENSE file for details.

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines on how to contribute to this project.

Support

If you need help or ha, please check our SUPPORT.md guide or open an issue on GitHub.

Code of Conduct

This project adheres to the Contributor Co Code of Conduct. Please read CODE_OF_CONDUCT.md for details.

项目信息

开发语言：Python

Star 数：72

技术要点

该项目属于CTF思路方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕Training Language Model Agents to Find Vulnerabilities with CTF-Dojo展开，提供了可运行的实现，对理解相关安全机制、工具原理与攻防技术具有较高的研究价值。

从技术实现角度看，此类项目的核心在于对底层原理的深入理解与工程化落地。开发者在研究过程中可以结合官方文档与社区讨论，梳理项目的初始化流程、核心模块职责与数据流转，逐步掌握其中的技术细节与设计思路。

使用与学习建议

1. 下载源码压缩包，解压到本地环境。

2. 根据项目类型准备对应的运行环境与依赖。

3. 按照 README 中的步骤完成配置与部署。

4. 在隔离的测试环境中运行验证，避免影响生产系统。

免责声明

本资源来源于开源社区，仅供安全研究、技术学习与合法授权范围内的攻防演练使用，请勿用于任何非法或侵害他人权益的用途。使用本项目产生的一切后果由使用者自行承担。