Academic Press/Elsevier, 2023. — 300 p. — ISBN: 978-0-12-824020-5.
Adversarial Robustness for Machine Learning summarizes the recent progress on this topic and introduces popular algorithms for the adversarial attack, defense, and verification. Sections cover the adversarial attack, assurance, and defense, mainly focusing on image classification applications which are the standard benchmark considered in the adversarial robustness community. Other sections discuss adversarial examples beyond image classification, other threat models beyond testing time attacks, and applications on adversarial robustness. For researchers, this book provides a thorough literature review that summarizes the latest progress in the area, which can be a good reference for conducting future research.
In addition, the book can also be used as a textbook for graduate courses on adversarial robustness or trustworthy Machine Learning. While Machine Learning (ML) algorithms have achieved remarkable performance in many applications, recent studies have demonstrated their lack of robustness against adversarial disturbance. The lack of robustness brings security concerns in ML models for real applications such as self-driving cars, robotics controls, and healthcare systems.
Summarizes the whole field of adversarial robustness for Machine learning models.
Provides a clearly explained, self-contained reference.
Introduces formulations, algorithms, and intuitions.
PreliminariesBackground and motivation.
Adversarial attackWhite-box adversarial attacks.
Black-box adversarial attacks.
Physical adversarial attacks.
Training-time adversarial attacks.
Adversarial attacks beyond image classification.
Robustness verificationOverview of neural network verification.
Incomplete neural network verification.
Complete neural network verification.
Verification against semantic perturbations.
Adversarial defenseOverview of adversarial defense.
Adversarial training.
Randomization-based defense.
Certified robustness training.
Adversary detection.
Adversarial robustness of beyond neural network models.
Adversarial robustness in meta-learning and contrastive learning.
Applications beyond attack and defenseModel reprogramming.
Contrastive explanations.
Model watermarking and fingerprinting.
Data augmentation for unsupervised machine learning.