本课程聚焦LLM与生成式AI安全，系统讲解OWASP GenAI十大风险、提示词注入、RAG与向量数据库防护，并涵盖红队测试与纵深防御，助你实战加固AI应用。
原始标题：LLM Security Made Simple: OWASP Top 10 for GenAI
这门专家级在线课程是纯粹的 AI 安全（AI Security）实战教程，核心围绕生成式 AI 与大语言模型（LLM）应用的漏洞攻防与安全架构设计展开。课程紧跟 AI 技术从“聊天机器人”向“自主执行任务的 AI Agent（智能体）”演进的趋势，全面剖析大模型应用独有的安全边界。学员将深度学习最权威的 OWASP GenAI 十大安全风险，掌握如何防范提示词注入（Prompt Injection）、系统提示词泄露、敏感数据泄露以及模型投毒等前沿新型攻击。
课程的核心价值在于“攻防兼备”，手把手带学员保护 RAG 管道、向量数据库及 MCP 集成等现代 AI 核心基础设施。你不仅能学会利用护栏（Guardrails）、访问控制和威胁建模来构建纵深防御体系，还能掌握黑客视角的 AI 红队测试（Red Teaming） 技术；最终通过一个完整的实战终期项目，亲自对一个充满漏洞的生成式 AI 应用进行安全加固，非常适合希望进阶 AI 架构安全的开发者与网络安全从业者。
Published 10/2026
Created by Medhat Gadallah
MP4 | Video: h264, 1280×720 | Audio: AAC, 44.1 KHz, 2 Ch
Level: Expert | Genre: eLearning | Language: English | Duration: 22 Lectures ( 3h 53m ) | Size: 2.7 GB
Master Prompt Injection, RAG Security, AI Agents, Red Teaming, and Secure Generative AI Application Design
What you’ll learn
⚡ Understand how LLM and Generative AI applications work and where their major security boundaries exist.
⚡ Identify the OWASP Top 10 security risks affecting LLM and Generative AI applications.
⚡ Recognize and mitigate Prompt Injection, sensitive data exposure, poisoning, unsafe output handling, and excessive AI agency.
⚡ Secure RAG pipelines, vector databases, AI agents, tools, MCP integrations, prompts, and sensitive data.
⚡ Apply threat modeling, guardrails, access controls, monitoring, and defense-in-depth to Generative AI systems.
⚡ – Test and evaluate LLM applications using practical AI security testing and red-team techniques.
Requirements
❗ No previous LLM security experience is required.
❗ Basic computer and internet knowledge is enough to start.
❗ Basic understanding of software applications is helpful but not mandatory.
❗ Cybersecurity experience is helpful, but beginners are welcome.
❗ Basic programming knowledge can help with practical examples but is not required to understand the core concepts.
❗ – Most importantly, bring curiosity about AI, cybersecurity, and how modern Generative AI applications can be protected.
Description
This course contains the use of artificial intelligence
Generative AI is rapidly moving from simple chatbots into real applications that can search company knowledge, call tools, access databases, generate code, interact with APIs, and even perform actions autonomously.
That creates an entirely new security challenge.
LLM Security Made Simple: OWASP Top 10 for GenAI is a beginner-friendly practical course designed to help you understand how modern Generative AI systems can fail, how attackers can abuse them, and how developers and security professionals can build stronger defenses.
You will begin by understanding how Large Language Models and LLM-powered applications actually work, including prompts, system instructions, APIs, RAG, embeddings, vector databases, tools, and AI agents.
From there, you will explore theOWASP Top 10 for LLM and Generative AI applications, including Prompt Injection, Sensitive Information Disclosure, Supply Chain risks, Data and Model Poisoning, Improper Output Handling, Excessive Agency, System Prompt Leakage, Vector and Embedding Weaknesses, Misinformation, and Unbounded Consumption.
Throughout the course, you will learn about
✨ LLM and Generative AI application architecture
✨ AI threat modeling and attack surfaces
✨ Prompt Injection and indirect Prompt Injection
✨ Sensitive data and credential leakage
✨ AI supply-chain security
✨ Data and model poisoning
✨ Secure handling of LLM-generated output
✨ AI agents and excessive permissions
✨ System Prompt Leakage
✨ RAG and vector database security
✨ Hallucinations and misinformation
✨ Token abuse and Denial-of-Wallet attacks
✨ AI agents, tools, and MCP security
✨ LLM security testing and AI red teaming
✨ Input and output guardrails
✨ Authentication, authorization, and secrets
✨ Logging and security monitoring
✨ Secure AI development lifecycle
✨ Defense-in-depth architecture
✨ Practical security exercises and a final capstone project
The course does not treat AI security as simply adding a filter around a chatbot. Instead, you will learn how security applies across data, prompts, models, retrieval, tools, permissions, outputs, infrastructure, and monitoring.
You will also bring everything together in a practical capstone where you review and secure a vulnerable Generative AI application. The curriculum specifically culminates in attack-surface mapping, threat modeling, securing RAG and agent permissions, implementing controls, red teaming, and producing a security assessment.
No previous LLM security knowledge is required. Concepts are introduced gradually, with practical examples and security scenarios designed to make difficult topics easier to understand.
Whether you are a developer, cybersecurity learner, AI engineer, AppSec professional, or simply interested in understanding the security risks behind modern Generative AI systems, this course will give you a structured foundation for continuing your journey into AI security.
Start learning how to design, test, and defend more secure Generative AI applications.
Who this course is for
⭐ Beginners interested in AI and cybersecurity.
⭐ Cybersecurity students who want to understand LLM and Generative AI security.
⭐ Software developers building AI-powered applications.
⭐ AI and LLM developers who want to build more secure systems.
⭐ Application Security and cybersecurity professionals expanding into AI security.
⭐ Prompt engineers working with LLM-powered systems.
⭐ Developers building RAG applications and AI agents.
⭐ Technical professionals who need to understand the risks of adopting Generative AI.
⭐ – Hobby learners interested in understanding how AI applications can be attacked and defended.
此处内容需要权限查看
会员免费查看
