Search icon CANCEL
Arrow left icon
Explore Products
Best Sellers
New Releases
Books
Videos
Audiobooks
Learning Hub
Conferences
Free Learning
Arrow right icon
Arrow up icon
GO TO TOP
Adversarial AI Attacks, Mitigations, and Defense Strategies

You're reading from   Adversarial AI Attacks, Mitigations, and Defense Strategies A cybersecurity professional's guide to AI attacks, threat modeling, and securing AI with MLSecOps

Arrow left icon
Product type Paperback
Published in Jul 2024
Publisher Packt
ISBN-13 9781835087985
Length 586 pages
Edition 1st Edition
Languages
Arrow right icon
Author (1):
Arrow left icon
John Sotiropoulos John Sotiropoulos
Author Profile Icon John Sotiropoulos
John Sotiropoulos
Arrow right icon
View More author details
Toc

Table of Contents (27) Chapters Close

Preface 1. Part 1: Introduction to Adversarial AI FREE CHAPTER
2. Chapter 1: Getting Started with AI 3. Chapter 2: Building Our Adversarial Playground 4. Chapter 3: Security and Adversarial AI 5. Part 2: Model Development Attacks
6. Chapter 4: Poisoning Attacks 7. Chapter 5: Model Tampering with Trojan Horses and Model Reprogramming 8. Chapter 6: Supply Chain Attacks and Adversarial AI 9. Part 3: Attacks on Deployed AI
10. Chapter 7: Evasion Attacks against Deployed AI 11. Chapter 8: Privacy Attacks – Stealing Models 12. Chapter 9: Privacy Attacks – Stealing Data 13. Chapter 10: Privacy-Preserving AI 14. Part 4: Generative AI and Adversarial Attacks
15. Chapter 11: Generative AI – A New Frontier 16. Chapter 12: Weaponizing GANs for Deepfakes and Adversarial Attacks 17. Chapter 13: LLM Foundations for Adversarial AI 18. Chapter 14: Adversarial Attacks with Prompts 19. Chapter 15: Poisoning Attacks and LLMs 20. Chapter 16: Advanced Generative AI Scenarios 21. Part 5: Secure-by-Design AI and MLSecOps
22. Chapter 17: Secure by Design and Trustworthy AI 23. Chapter 18: AI Security with MLSecOps 24. Chapter 19: Maturing AI Security 25. Index 26. Other Books You May Enjoy

Supply-chain attacks in LLMs

So far, we assumed the use of SaaS-hosted proprietary pre-trained models accessed via their API. This was the main setting at the beginning of the LLM explosion in 2023, and vendors such as Anthropic, Google, and especially OpenAI have dominated the scene.

However, this has changed, and there has been an explosive growth of open-access models.

Note

Open access refers to the ability to download and use or redistribute a model. In many articles and discussions, the term open source models is used. However, since no source code has been released, the term open access is more appropriate.

A variant of open-access models is open-weight models, where the models’ weights are publicly available for research, development, and application purposes. These models are open because the community can access, use, modify, and distribute the model weights without restrictive licenses.

In addition to Meta’s Llama, the first open-access LLM, new...

lock icon The rest of the chapter is locked
Register for a free Packt account to unlock a world of extra content!
A free Packt account unlocks extra newsletters, articles, discounted offers, and much more. Start advancing your knowledge today.
Unlock this book and the full library FREE for 7 days
Get unlimited access to 7000+ expert-authored eBooks and videos courses covering every tech area you can think of
Renews at $19.99/month. Cancel anytime