Learning objectives

Contents

• Introduction Introduction

  • Securing generative AI: Introduction

    2m 59s
• 1. Introduction to AI Threats and LLM Security 1. Introduction to AI Threats and LLM Security

  • (Locked)
    Learning objectives

    1m 18s
  • (Locked)
    Understanding the significance of LLMs in the AI landscape

    7m 6s
  • Exploring the resources for this course: GitHub repositories and others

    2m 54s
  • (Locked)
    Introducing retrieval augmented generation (RAG)

    12m 24s
  • (Locked)
    Understanding the OWASP Top 10 risks for LLMs

    5m 46s
  • (Locked)
    Exploring the MITRE ATLAS™ (adversarial threat landscape for artificial intelligence systems) framework

    5m 38s
  • (Locked)
    Understanding the NIST taxonomy and terminology of attacks and mitigations

    7m 8s
• 2. Understanding Prompt Injection Insecure Output Handling 2. Understanding Prompt Injection Insecure Output Handling

  • (Locked)
    Learning objectives

    1m 1s
  • Defining prompt injection attacks

    11m 41s
  • (Locked)
    Exploring real-life prompt injection attacks

    3m 57s
  • (Locked)
    Using ChatML for OpenAI API calls to indicate to the LLM the source of prompt input

    10m 4s
  • (Locked)
    Enforcing privilege control on LLM access to back-end systems

    6m 10s
  • (Locked)
    Best practices around API tokens for plugins, data access, and function-level permissions

    3m 2s
  • (Locked)
    Understanding insecure output handling attacks

    3m 22s
  • (Locked)
    Using the OWASP ASVS to protect against insecure output handling

    4m 43s
• 3. Training Data Poisoning, Model Denial of Service Supply Chain Vulnerabilities 3. Training Data Poisoning, Model Denial of Service Supply Chain Vulnerabilities

  • (Locked)
    Learning objectives

    46s
  • Understanding training data poisoning attacks

    4m 26s
  • (Locked)
    Exploring model denial of service attacks

    3m 11s
  • (Locked)
    Understanding the risks of the AI and ML supply chain

    8m 33s
  • (Locked)
    Best practices when using open-source models from Hugging face and other sources

    12m 45s
  • (Locked)
    Securing Amazon Bedrock, Amazon SageMaker, Microsoft Azure AI services, and other environments

    16m 5s
• 4. Sensitive Information Disclosure, Insecure Plugin Design, and Excessive Agency 4. Sensitive Information Disclosure, Insecure Plugin Design, and Excessive Agency

  • (Locked)
    Learning objective

    1m 4s
  • (Locked)
    Understanding sensitive information disclosure

    2m 52s
  • (Locked)
    Exploiting insecure plugin design

    3m 11s
  • (Locked)
    Avoiding excessive agency

    3m 46s
• 5. Overreliance, Model Theft, and Red Teaming AI Models 5. Overreliance, Model Theft, and Red Teaming AI Models

  • (Locked)
    Learning objectives

    47s
  • (Locked)
    Understanding overreliance

    5m 17s
  • (Locked)
    Exploring model theft attacks

    3m 49s
  • (Locked)
    Understanding red teaming of AI models

    13m 33s
• 6. Protecting Retrieval Augmented Generation (RAG) Implementations 6. Protecting Retrieval Augmented Generation (RAG) Implementations

  • (Locked)
    Learning objectives

    1m 5s
  • (Locked)
    Understanding RAG, LangChain, Llama index, and AI orchestration

    17m 21s
  • (Locked)
    Securing embedding models

    9m 38s
  • (Locked)
    Securing vector databases

    12m
  • (Locked)
    Monitoring and incident response

    7m 50s
• Conclusion Conclusion

  • (Locked)
    Securing generative AI: Summary

    1m 41s