A.I.G (AI-Infra-Guard) 集成AI基础设施漏洞扫描、MCP Server风险检测与大模型安全体检等能力,旨在为用户提供最全面、智能与易用的AI安全风险自查解决方案。
我们致力于将A.I.G(AI-Infra-Guard)打造为业界领先的 AI 红队工具平台。更多的 Star 能让这个项目被更多人看到,吸引更多的开发者参与进来,从而让项目更快地迭代和完善。您的 Star 对我们至关重要!
📢 News:
| Docker | 内存 | 磁盘空间 |
|---|---|---|
| 20.10 或更高 | 4GB+ | 10GB+ |
# 此方法会从 Docker Hub 拉取预构建的镜像
git clone https://github.com/Tencent/AI-Infra-Guard.git
cd AI-Infra-Guard
# Docker Compose V2+ 请将 'docker-compose' 替换为 'docker compose'
docker-compose -f docker-compose.images.yml up -d服务启动后,您可以通过以下地址访问 A.I.G 的 Web 界面:
http://localhost:8088
📦 更多安装方式
方式 2:一键安装脚本(推荐)
# 此方法将自动安装 Docker 并启动 A.I.G
curl https://raw.githubusercontent.com/Tencent/AI-Infra-Guard/refs/heads/main/docker.sh | bash方式 3:源码编译运行
git clone https://github.com/Tencent/AI-Infra-Guard.git
cd AI-Infra-Guard
# 此方法从本地源代码构建 Docker 镜像并启动服务
# (Docker Compose V2+ 请将 'docker-compose' 替换为 'docker compose')
docker-compose up -d注意:AI-Infra-Guard 项目定位为企业或个人内部使用的 AI 红队测试平台,目前暂无鉴权认证机制,请勿在公网环境中部署使用。
更多信息请参阅:https://tencent.github.io/AI-Infra-Guard/?menu=getting-started
体验具有内测及高级功能的Pro版,需要邀请码,优先提供给提交过 Issues、Pull Requests 或 Discussions,或积极帮助社区发展的贡献者。访问:https://aigsec.ai/
| 特性 | 详细信息 |
|---|---|
| AI基础设施漏洞扫描 | 精准识别30+AI框架组件,覆盖400+已知CVE漏洞,包括Ollama/ComfyUI/vLLM等 |
| AI工具协议扫描 | 基于AI Agent驱动,检测14大类MCP Server与Skills安全风险,支持源代码/远程URL扫描 |
| 大模型安全体检 | 快速评估Prompt安全风险,内置多个精选越狱评估数据集,支持多种越狱攻击算子,跨模型安全性能对比 |
其他优势
- 🖥️ 现代化Web界面:用户友好的UI,一键扫描和实时进度跟踪
- 🔌 完整API:完整的接口文档和Swagger规范,便于集成
- 🌐 多语言支持:中英文界面,本地化文档
- 🐳 跨平台兼容:支持Linux、macOS和Windows,基于Docker部署
- 🆓 免费且开源:完全免费,MIT开源协议
访问我们的在线文档:https://tencent.github.io/AI-Infra-Guard/
更多详细的常见问题解答和故障排除指南,请访问我们的文档。
系统架构演进说明:docs/architecture_versions.md。
A.I.G 提供了一套创建任务相关的API接口,支持AI基础设施扫描、MCP安全扫描和大模型安全体检功能。
项目运行后访问 http://localhost:8088/docs/index.html 可查看完整的API文档
详细的API使用说明、参数说明和完整示例代码,请查看 完整API文档。
A.I.G 的核心能力之一就是其丰富且可快速配置的插件系统。我们欢迎社区贡献高质量的插件和功能。
- 指纹规则: 在
data/fingerprints/目录下添加新的 YAML 指纹文件 - 漏洞规则: 在
data/vuln/目录下添加新的漏洞检测规则 - MCP 插件: 在
data/mcp/目录下添加新的 MCP 安全检测规则 - 模型评测集: 在
data/eval目录下添加新的模型评测集
请参考现有规则格式,创建新文件并通过 Pull Request 提交。
我们诚挚感谢学术合作伙伴提供的卓越研究协作。
|
李挥教授 |
王滨 |
刘泽心 |
余昊 |
杨傲 |
林郑熹 |
|
杨哲慜教授 |
钟康维 |
林佳鹏 |
盛铖 |
![]() |
![]() |
| 微信群 | Discord [链接] |
|---|---|
![]() |
![]() |
如有合作咨询或反馈,请联系我们:zhuque@tencent.com
如果你对代码安全感兴趣,推荐关注腾讯悟空代码安全团队开源的行业首个项目级 AI 生成代码安全性评测框架A.S.E(AICGSecEval)。
如果您在研究或产品中使用了A.I.G,请使用以下引用:
@misc{Tencent_AI-Infra-Guard_2025,
author={{Tencent Zhuque Lab}},
title={{AI-Infra-Guard: A Comprehensive, Intelligent, and Easy-to-Use AI Red Teaming Platform}},
year={2025},
howpublished={GitHub repository},
url={https://github.com/Tencent/AI-Infra-Guard}
}我们深深感谢在学术工作中使用A.I.G,并为推进AI安全研究做出贡献的团队:
[1] Ruiqi Li, Zhiqiang Wang, Yunhao Yao et al. "MCP-ITP: An Automated Framework for Implicit Tool Poisoning in MCP." arXiv preprint arXiv:2601.07395v1 (2026). [pdf]
[2] Jingxiao Yang, Ping He, Tianyu Du et al. "HogVul: Black-box Adversarial Code Generation Framework Against LM-based Vulnerability Detectors." arXiv preprint arXiv:2601.05587v1 (2026). [pdf]
[3] Yunyi Zhang, Shibo Cui, Baojun Liu et al. "Beyond Jailbreak: Unveiling Risks in LLM Applications Arising from Blurred Capability Boundaries." arXiv preprint arXiv:2511.17874v2 (2025). [pdf]
[4] Teofil Bodea, Masanori Misono, Julian Pritzi et al. "Trusted AI Agents in the Cloud." arXiv preprint arXiv:2512.05951v1 (2025). [pdf]
[5] Christian Coleman. "Behavioral Detection Methods for Automated MCP Server Vulnerability Assessment." [pdf]
[6] Bin Wang, Zexin Liu, Hao Yu et al. "MCPGuard : Automatically Detecting Vulnerabilities in MCP Servers." arXiv preprint arXiv:22510.23673v1 (2025). [pdf]
[7] Weibo Zhao, Jiahao Liu, Bonan Ruan et al. "When MCP Servers Attack: Taxonomy, Feasibility, and Mitigation." arXiv preprint arXiv:2509.24272v1 (2025). [pdf]
[8] Ping He, Changjiang Li, et al. "Automatic Red Teaming LLM-based Agents with Model Context Protocol Tools." arXiv preprint arXiv:2509.21011 (2025). [pdf]
[9] Yixuan Yang, Daoyuan Wu, Yufan Chen. "MCPSecBench: A Systematic Security Benchmark and Playground for Testing Model Context Protocols." arXiv preprint arXiv:2508.13220 (2025). [pdf]
[10] Zexin Wang, Jingjing Li, et al. "A Survey on AgentOps: Categorization, Challenges, and Future Directions." arXiv preprint arXiv:2508.02121 (2025). [pdf]
[11] Yongjian Guo, Puzhuo Liu, et al. "Systematic Analysis of MCP Security." arXiv preprint arXiv:2508.12538 (2025). [pdf]
📧 如果您在研究中使用了A.I.G,请联系我们,让更多人看到您的研究!
本项目基于 MIT 许可证 开源。详细信息请查阅 License.txt 文件。










