Samuel Sambasivam: Federated Learning for Privacy-Preserving AI Systems, Gebunden

Klappentext

This book presents a rigorous and comprehensive treatment of federated learning as a foundational paradigm for privacy-preserving artificial intelligence. Integrating theoretical principles with implementation strategies and domain-specific applications, it offers a unified framework for understanding the design, optimization, and deployment of distributed AI systems in privacy-sensitive environments.

The volume systematically examines the architectures and operational models of horizontal, vertical, cross-device, and cross-silo federated learning. Core optimization algorithms---including FedAvg, FedProx, personalized federated learning methods, and asynchronous federated approaches---are analyzed in detail, with particular attention to convergence behavior under non-IID and heterogeneous data distributions. The text further explores the mathematical and systems foundations that enable secure and trustworthy collaboration across decentralized environments.

A substantial portion of the book is devoted to privacy-preserving and security-enhancing mechanisms that underpin modern federated systems. Topics include differential privacy, secure aggregation, homomorphic encryption, secure multi-party computation, Byzantine-resilient aggregation, and adversarial robustness. These techniques are evaluated not only from a theoretical perspective but also in terms of their practical implications for scalability, communication efficiency, model utility, and deployment.

To bridge theory and practice, the book presents detailed application studies in financial systems, cybersecurity for zero-day attack detection, and healthcare diagnostics. Each case study includes experimental design, dataset considerations, baseline comparisons, implementation workflows, performance evaluation, and critical discussion of practical challenges and research opportunities. A dedicated design science chapter further guides readers through requirements analysis, system architecture, deployment strategies, and operational best practices for enterprise-scale federated AI systems.

Designed for graduate students, researchers, and industry practitioners, this text provides a pedagogically integrated resource that combines analytical rigor with practical relevance. Readers will benefit from worked examples, implementation guidance, comparative analyses, and end-of-chapter exercises that support both academic study and real-world application. By unifying theoretical foundations, privacy-preserving methodologies, and production-oriented considerations within a single volume, this book serves as an authoritative reference for the next generation of secure and decentralized AI systems.