The Advancements and Future Prospects of Federated Learning-Based Methods for Biometrics

Yifan Zhang

2024

Abstract

Biometrics has gained great attention due to its effectiveness in security applications recently. Although Deep Learning (DL) has proven useful in biometrics, but it requires huge, high-quality datasets and raises privacy issues. Federated Learning (FL) provides a solution, which trains models on clients and sending updates to the global model without sharing sensitive information. This paper reviews recent advances in applying FL for biometrics, mainly focusing on its use in face recognition, iris recognition, palmprint recognition, and finger vein recognition. Some FL-based methods such as FedFace, FedGC, PrivacyFace, and other methods are introduced, and their contributions are discussed in this paper. These methods have greatly contributed to privacy preservation and model performance. Despite the great progress, there are still great challenges and limitations in handling model interpretability and applicability due to the non-independent and identically distributed (non-IID) data and model complexity. The future prospects include enhancing model interpretability through techniques such as SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) and improving applicability by implementing transfer learning and domain adaptation. This paper provides suggestions and references for future research and concludes that FL provides a promising path forward for biometrics.

Paper Citation

in Harvard Style

Zhang Y. (2024). The Advancements and Future Prospects of Federated Learning-Based Methods for Biometrics. In Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML; ISBN 978-989-758-754-2, SciTePress, pages 423-427. DOI: 10.5220/0013525300004619

in Bibtex Style

@conference{daml24,
author={Yifan Zhang},
title={The Advancements and Future Prospects of Federated Learning-Based Methods for Biometrics},
booktitle={Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML},
year={2024},
pages={423-427},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013525300004619},
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 - The Advancements and Future Prospects of Federated Learning-Based Methods for Biometrics
SN - 978-989-758-754-2
AU - Zhang Y.
PY - 2024
SP - 423
EP - 427
DO - 10.5220/0013525300004619
PB - SciTePress