Multimodal Biometric Authentication Method By Federated Learning

Abstract

Biometric authentication systems, which use unique biological and behavioral characteristics such as
fingerprints, facial recognition, iris patterns, and voice recognition, have become increasingly prevalent in secure
access control applications. However, the widespread adoption of these systems raises significant concerns about
data privacy, security, and scalability, particularly when sensitive biometric data is centrally stored and
processed. To address these challenges, this study introduces a federated learning-based framework for
multimodal biometric authentication. Federated learning enables decentralized model training across multiple
devices or nodes, ensuring that raw biometric data remains localized and never shared with central servers. This
preserves user privacy while allowing the system to learn from distributed data. The proposed approach integrates
multiple biometric modalities to enhance authentication accuracy, leveraging complementary information from
different data sources. Advanced deep learning models are employed to extract and fuse features from these
modalities, ensuring robustness against variations in data quality and environmental conditions. The framework
addresses challenges such as data heterogeneity, communication constraints, and device resource limitations
through techniques like differential privacy, secure aggregation, and model compression. Experimental
evaluations are conducted using real-world multimodal biometric datasets to assess the system's performance.
Results demonstrate improved authentication accuracy and robustness compared to unimodal and traditional
centralized systems. Furthermore, the federated approach significantly reduces privacy risks and ensures
compliance with data protection regulations. This work highlights the potential of federated learning in
developing secure, scalable, and privacy-preserving biometric systems, paving the way for its application in
diverse domains such as mobile security, healthcare, and financial services. The findings underscore the
importance of combining federated learning with multimodal biometrics to achieve the next generation of reliable
and user-centric authentication methods.