人工智能安全学习中心
结构化的学习路径、认证和社区资源可帮助您构建 AI 安全专业知识 - 无论您是红队成员、开发人员、分析师还是合规专业人士。
技能水平
初学者
刚刚开始接触人工智能安全?从这里开始。
- AI 安全术语表 — 学习术语
- RAG 特定安全测试框架 — 了解风险状况
- 快速注射基础知识 — #1 漏洞解释
- 测试方法概述 — AI 渗透测试的工作原理
预计。时间: 2-3 weeks
中级
准备好深入研究技术和工具。
- RAG 安全深入探讨 — 矢量数据库攻击、中毒
- MCP 安全指南 — 模型上下文协议漏洞
- AI渗透测试工具 — Garak、PyRIT、LLM Guard 实践
- 安全检查表 — 应用结构化测试
预计。时间: 1-2 months
职业学习路径
AH
Frequently Asked Questions
How do I start learning AI security?
Start with foundational knowledge of LLM architecture and common vulnerabilities (OWASP LLM Top 10). Practice with hands-on labs like OWASP's AI security challenges, Hack The Box's AI modules, and local LLM testing with tools like Ollama. Follow structured learning paths that cover prompt injection, red teaming, and secure AI development.
What are the best AI security certifications available?
Top AI security certifications include: AI Security Certified Professional (AISCP), Certified AI Security Professional (CAISP), CEH with AI security specialization, and SANS AI security courses. These certifications cover LLM security testing, adversarial machine learning, and AI risk management frameworks.
What learning path should I follow for AI red teaming?
An AI red teaming learning path includes: 1) Master prompt injection techniques and jailbreaking, 2) Learn MCP server security testing, 3) Study model extraction and inversion attacks, 4) Practice with tools like Garak and PyRIT, 5) Understand agentic AI vulnerabilities and OWASP Agentic Top 10.
What are the best resources for learning LLM security?
Best LLM security resources include: OWASP LLM Top 10 and Agentic Top 10 guides, AI Hacking Portal's comprehensive testing methodology, NIST AI RMF documentation, MITRE ATLAS knowledge base, and Capture The Flag challenges like the OWASP FinBot CTF for hands-on agentic AI security practice.