Federated Learning in Customer-Centric Applications: Balancing Privacy, Personalization and Performance

Xinxiang Gao

2024

Abstract

Federated Learning (FL) offers a practical way to meet the growing need for privacy-preserving machine learning, especially in customer-focused areas like finance, retail, and business collaboration. It allows models to be trained securely and in a decentralized way, without collecting sensitive data in one place. This paper examines how FL facilitates secure and decentralized model training while avoiding the centralization of sensitive data. It addresses key challenges such as data heterogeneity, model convergence, and privacy concerns, proposing methods like clustered FL, asynchronous updates, and attention-based mechanisms for improving model performance. The paper also discusses privacy vulnerabilities, such as gradient leakage, and explores solutions like differential privacy and secure aggregation. Although federated learning enhances data privacy and service personalization, it faces limitations, including increased computational complexity and communication overhead. Future research needs to prioritize enhancing privacy safeguards while ensuring model accuracy and scalability. This review can serve as a valuable reference for researchers looking to understand the advancements in this field.

Paper Citation

in Harvard Style

Gao X. (2024). Federated Learning in Customer-Centric Applications: Balancing Privacy, Personalization and Performance. In Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML; ISBN 978-989-758-754-2, SciTePress, pages 537-542. DOI: 10.5220/0013527900004619

in Bibtex Style

@conference{daml24,
author={Xinxiang Gao},
title={Federated Learning in Customer-Centric Applications: Balancing Privacy, Personalization and Performance},
booktitle={Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML},
year={2024},
pages={537-542},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013527900004619},
isbn={978-989-758-754-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML
TI - Federated Learning in Customer-Centric Applications: Balancing Privacy, Personalization and Performance
SN - 978-989-758-754-2
AU - Gao X.
PY - 2024
SP - 537
EP - 542
DO - 10.5220/0013527900004619
PB - SciTePress