Home // ADVCOMP 2024, The Eighteenth International Conference on Advanced Engineering Computing and Applications in Sciences // View article

Accelerating Differential Privacy-Based Federated Learning Systems

Authors:
Mirco Mannino
Alessio Medaglini
Biagio Peccerillo
Sandro Bartolini

Keywords: differential privacy; federated learning; hardware accelerator.

Abstract:
The number of mobile, wearable, and Internet of Things (IoT) devices we are using is increasingly growing, especially those implementing machine learning applications on-the-edge. Relying on a centralized server for processing and storing of this ever-increasing amount of data might not be the optimal solution, from both performance and privacy points of view. Federated Learning is a good solution to avoid sending user’s data to a central server to train machine learning models. In order to guarantee privacy in a Federated Learning system, it is possible to leverage several techniques. Differential Privacy is one of the most popular, since it provides robust privacy protection. In this paper, we target mobile devices, proposing ideas on how to speed up training in Differential Privacy-based Federated Learning systems through a dedicated hardware accelerator.

Pages: 8 to 10

Copyright: Copyright (c) IARIA, 2024

Publication date: September 29, 2024

Published in: conference

ISSN: 2308-4499

ISBN: 978-1-68558-184-8

Location: Venice, Italy

Dates: from September 29, 2024 to October 3, 2024