Adversarial AI Attacks, Mitigations, and Defense Strategies - John Sotiropoulos

Learn how to defend AI and LLM systems against manipulation and intrusion through adversarial attacks such as poisoning, trojan horses, and model extraction, leveraging DevSecOps, MLOps and other methods to secure systems

Key Features:

- Understand the unique security challenges presented by predictive and generative AI

- Explore common adversarial attack strategies as well as emerging threats such as prompt injection

- Mitigate the risks of attack on your AI system with threat modeling and secure-by-design methods

- Purchase of the print or Kindle book includes a free PDF eBook

Book Description:

Adversarial attacks trick AI systems with malicious data, creating new security risks by exploiting how AI learns. This challenges cybersecurity as it forces us to defend against a whole new kind of threat. This book demystifies adversarial attacks and equips you with the skills to secure AI technologies, moving beyond research hype or business-as-usual activities.

This strategy-based book is a comprehensive guide to AI security, presenting you with a structured approach with practical examples to identify and counter adversarial attacks. In Part 1, you'll touch on getting started with AI and learn about adversarial attacks, before Parts 2, 3 and 4 move through different adversarial attack methods, exploring how each type of attack is performed and how you can defend your AI system against it. Part 5 is dedicated to introducing secure-by-design AI strategy, including threat modeling and MLSecOps and consolidating recent research, industry standards and taxonomies such as OWASP and NIST. Finally, based on the classic NIST pillars, the book provides a blueprint for maturing enterprise AI security, discussing the role of AI security in safety and ethics as part of Trustworthy AI.

By the end of this book, you'll be able to develop, deploy, and secure AI systems against the threat of adversarial attacks effectively.

What You Will Learn:

- Set up a playground to explore how adversarial attacks work

- Discover how AI models can be poisoned and what you can do to prevent this

- Learn about the use of trojan horses to tamper with and reprogram models

- Understand supply chain risks

- Examine how your models or data can be stolen in privacy attacks

- See how GANs are weaponized for Deepfake creation and cyberattacks

- Explore emerging LLM-specific attacks, such as prompt injection

- Leverage DevSecOps, MLOps and MLSecOps to secure your AI system

Who this book is for:

This book tackles AI security from both angles - offense and defence. AI developers and engineers will learn how to create secure systems, while cybersecurity professionals, such as security architects, analysts, engineers, ethical hackers, penetration testers, and incident responders will discover methods to combat threats to AI and mitigate the risks posed by attackers. The book also provides a secure-by-design approach for leaders to build AI with security in mind. To get the most out of this book, you'll need a basic understanding of security, ML concepts, and Python.

Table of Contents

- Getting Started with AI

- Building Our Adversarial Playground

- Security and Adversarial AI

- Poisoning Attacks

- Model Tampering with Trojan Horses and Model Reprogramming

- Supply Chain Attacks and Adversarial AI

- Evasion Attacks against Deployed AI

- Privacy Attacks - Stealing Models

- Privacy Attacks - Stealing Data

(N.B. Please use the Read Sample option to see further chapters)