Research on Privacy Protection Technology in Federated Learning

Zihan Xiang

2025

Abstract

The extensive implementation of machine-learning techniques, the exponential expansion of big data, and the reinforcement of global legal provisions regarding data privacy safeguarding have spurred the swift advancement of federated learning. The primary benefit of federated learning is manifested in its capacity to carry out collaborative data training while refraining from the sharing of unprocessed data, which is crucial for protecting user privacy and complying with data protection regulations. This paper first summarizes the basic definition, classification, and algorithm principles of federated learning and then focuses on the applications of differential privacy and homomorphic encryption techniques within the privacy protection domain of federated learning. Differential privacy safeguards data privacy through the addition of noise when updating models. It yields a favorable outcome in terms of privacy protection, particularly within the medical sector. However, it encounters difficulties in achieving a balance between privacy protection and model accuracy, as well as in determining the value of the privacy budget. Homomorphic encryption enables direct calculations to be carried out on ciphertexts, achieving privacy protection throughout the entire process of federated learning. It has strong compatibility and wide applications but has high computational costs and low performance in large-scale distributed systems. In the future, privacy protection technologies for federated learning will develop towards multi-technology integration, adaptation to emerging scenarios, and standardization and normalization to address challenges such as inference attacks, data heterogeneity, and malicious attacks, promote the secure and compliant sharing of data, and facilitate the development of a digital society.

Paper Citation

in Harvard Style

Xiang Z. (2025). Research on Privacy Protection Technology in Federated Learning. In Proceedings of the 2nd International Conference on Data Science and Engineering - Volume 1: ICDSE; ISBN 978-989-758-765-8, SciTePress, pages 117-122. DOI: 10.5220/0013679600004670

in Bibtex Style

@conference{icdse25,
author={Zihan Xiang},
title={Research on Privacy Protection Technology in Federated Learning},
booktitle={Proceedings of the 2nd International Conference on Data Science and Engineering - Volume 1: ICDSE},
year={2025},
pages={117-122},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013679600004670},
isbn={978-989-758-765-8},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Data Science and Engineering - Volume 1: ICDSE
TI - Research on Privacy Protection Technology in Federated Learning
SN - 978-989-758-765-8
AU - Xiang Z.
PY - 2025
SP - 117
EP - 122
DO - 10.5220/0013679600004670
PB - SciTePress