Overview

The first practical field guide to adversarial AI hacking. AI models are everywhere - in self-driving cars, face recognition, voice assistants, spam filters, and fraud detection systems. But here's the uncomfortable truth: they can all be hacked. In this hands-on book, you'll learn how to attack and defend machine learning systems step by step. Through code labs and real-world case studies, you'll see how small changes can fool massive models - and how to build defenses that actually work. Inside you'll discover: - How to run evasion attacks (FGSM, PGD) that flip model predictions - How to perform poisoning attacks and build hidden backdoors - How to extract secrets with model inversion and membership inference - How to break vision, NLP, and speech models with real adversarial inputs - How to exploit multimodal models like CLIP with mismatched inputs - How to defend using adversarial training, defensive distillation, and input sanitization - How to build detection pipelines that flag attacks in real time - How to use toolkits like Foolbox, CleverHans, and TextAttack - The ethics, policy, and future of adversarial AI Who this book is for: Ethical hackers, ML engineers, cybersecurity pros, red teamers, and anyone who wants to stay ahead in the new arms race where AI attacks AI. Adversarial AI for Hackers doesn't just explain vulnerabilities - it gives you working code, labs, and a hacker's mindset. Don't just trust AI. Learn how to break it - and defend it.

Full Product Details

9798263584177

Table of Contents

Reviews

Author Information

Tab Content 6

Countries Available