Strengthening Deep Neural Networks: Making AI Less Susceptible to Adversarial Trickery

Attack vectors involving deep neural networks (DNNs) that can be deliberately fooled by data that wouldn't trick a human present a new threat. This book explores real-world scenarios where DNNs process image, audio, and video data and examines attack motivations, risks, and methods for increasing AI robustness. It is for data scientists, security architects, and anyone interested in the differences between artificial and biological perception.

Format: Paperback / softback
Length: 250 pages
Publication date: 09 August 2019
Publisher: O'Reilly Media, Inc, USA

As deep neural networks (DNNs) gain widespread adoption in real-world applications, the potential to exploit their vulnerabilities by presenting data that would not deceive a human emerges as a significant threat. This practical book delves into real-world scenarios where DNNs, the core algorithms underpinning much of artificial intelligence, are utilized daily to process image, audio, and video data. Author Katy Warr explores the motivations behind adversarial attacks, the risks associated with this adversarial input, and strategies for enhancing AI robustness against such attacks. Whether you are a data scientist developing DNN algorithms, a security architect seeking ways to improve the resilience of AI systems, or someone intrigued by the distinctions between artificial and biological perception, this book offers valuable insights.

Dive into the world of DNNs and uncover their susceptibility to being tricked by adversarial input. Explore the methods employed to generate adversarial input capable of fooling DNNs. Examine real-world scenarios and model the adversarial threat. Evaluate the robustness of neural networks and learn techniques to enhance the resilience of AI systems against adversarial data.

Furthermore, consider some potential ways in which AI may evolve to become even better at mimicking human perception in the years to come. By understanding the challenges and opportunities posed by deep neural networks, this book equips you with the knowledge and tools necessary to navigate the evolving landscape of AI security and develop more resilient and trustworthy systems.

Weight: 406g
Dimension: 233 x 178 x 13 (mm)
ISBN-13: 9781492044956