Artificial Intelligence: Data and Model Safety - Softcover

Jiang

Softcover

ISBN 10: 0443248400 ISBN 13: 9780443248405

Publisher: Elsevier, 2025

Synopsis

Artificial Intelligence Data and Model Safety: Risks, Attacks and Defenses offers a comprehensive overview of the evolution of AI and its security concerns. The book delves into how historical advancements in AI have both bolstered and complicated the issue of safeguarding data and models. By reflecting on the interplay between machine learning innovations and vulnerabilities, it sets the stage for readers to understand the critical importance of robust defenses in this era of digital and algorithmic reliance. In addition to contextualizing the historical trajectory of AI security, the book examines foundational elements of machine learning, emphasizing the mechanisms that contribute to, or mitigate, risks.

Readers are guided through case studies of real-world attacks, illustrating the practical implications of security weaknesses, while proposed defense strategies provide actionable insights for strengthening AI systems.

• Comprehensively introduces AI safety, covering both attack and defense technologies
• Covers a broad range of attack and defense strategies from the perspectives of adversarial learning and robust optimization, providing detailed explanations and insights
• Includes the latest research developments and state-of-the-art techniques in the field of AI security

About this title

About the Authors

Professor Yu-Gang Jiang is based at Fudan University, PR China. He is primarily engaged in scientific research in artificial intelligence, multimedia information processing, and secure and trustworthy machine learning. He has published over 100 papers in top international journals and conferences in these domains. In recent years, he has achieved multiple innovative results in artificial intelligence security, such as proposing the first black-box video adversarial sample generation method and the first data poisoning and backdoor attack methods for video recognition models.

Dr Xingjun Ma is an associate professor in the School of Computer Science and Technology, Fudan University, PR China. He obtained his doctoral degree from The University of Melbourne in Australia in 2019. He has previously worked as a research fellow at The University of Melbourne and as a lecturer at Deakin University. His research focuses on trustworthy machine learning, specifically the security, robustness, interpretability, privacy, and fairness of machine learning data, algorithms, and models. He has published over 50 papers in top international conferences and journals and holds two international patents.

Dr Zuxuan Wu is currently an assistant professor at the School of Computer Science and Technology, Fudan University, China. In 2020, he obtained his doctoral degree from the University of Maryland in the US. His main research interests include computer vision, deep learning, and multimedia content analysis. He has been awarded the AI 2000 Most Influential Scholars Award in 2022, and the Microsoft Research Ph.D. Fellowship in 2019, and the Snap Ph.D. Fellowship in 2017.

From the Back Cover

Artificial Intelligence Data and Model Safety: Risks, Attacks and Defenses begins with a brief review of the history of AI and AI security and then introduces the fundamental aspects of machine learning and AI security. Two key aspects are covered: data security and modelling. It provides detailed explanations of a wide range of attacks and defense algorithms related to data security, as well as adversarial attack/defense, backdoor attack/defense, and extraction attack/defense algorithms related to model security. By providing a systematic, comprehensive, and in-depth introduction to the topic, this book help readers understand the advanced attack and defense techniques in the field of AI security.

Edition details

• Publisher: Elsevier
• Publication date: 2025
• Language: English
• ISBN 10: 0443248400
• ISBN 13: 9780443248405
• Binding: Paperback
• Number of pages: 386