You're reading from Privacy-Preserving Machine Learning A use-case-driven approach to building and protecting ML pipelines from privacy and security threats

Product type Paperback

Published in May 2024

Publisher Packt

ISBN-13 9781800564671

Length 402 pages

Edition 1st Edition

Languages

Python

Tools

Federated Learning

Concepts

Machine Learning

Author (1):

Srinivasa Rao Aravilli

View More author details

Table of Contents (17) Chapters

Preface

1. Part 1: Introduction to Data Privacy and Machine Learning FREE CHAPTER

2. Chapter 1: Introduction to Data Privacy, Privacy Breaches, and Threat Modeling

3. Chapter 2: Machine Learning Phases and Privacy Threats/Attacks in Each Phase

4. Part 2: Use Cases of Privacy-Preserving Machine Learning and a Deep Dive into Differential Privacy

5. Chapter 3: Overview of Privacy-Preserving Data Analysis and an Introduction to Differential Privacy

6. Chapter 4: Overview of Differential Privacy Algorithms and Applications of Differential Privacy

7. Chapter 5: Developing Applications with Differential Privacy Using Open Source Frameworks

8. Part 3: Hands-On Federated Learning

9. Chapter 6: Federated Learning and Implementing FL Using Open Source Frameworks

10. Chapter 7: Federated Learning Benchmarks, Start-Ups, and the Next Opportunity

11. Part 4: Homomorphic Encryption, SMC, Confidential Computing, and LLMs

12. Chapter 8: Homomorphic Encryption and Secure Multiparty Computation

13. Chapter 9: Confidential Computing – What, Why, and the Current State

14. Chapter 10: Preserving Privacy in Large Language Models

15. Index

Why subscribe?

16. Other Books You May Enjoy

FL algorithms

FL algorithms, such as FedSGD, FedAvg, and Adaptive Federated Optimization, play a crucial role in the distributed training of ML models while ensuring privacy and security. In this section, we will explore these algorithms and their key characteristics.

FedSGD

Federated stochastic gradient descent (FedSGD) is a fundamental algorithm used in FL. It extends the traditional SGD optimization method to the federated setting. In FedSGD, each client (entity) computes the gradients on its local data and sends them to the central server. The server aggregates the gradients and updates the global model parameters accordingly. FedSGD is efficient for large-scale distributed training but may suffer from issues related to non-IID data and communication efficiency.