Training-free Dimensionality Reduction via Feature Truncation: Enhancing Efficiency in Privacy-preserving Multi-Biometric Systems

2025-08-15

AI Summary

To address the challenges of large template sizes and high computational overhead of homomorphic encryption in privacy-preserving multimodal biometric recognition, this paper proposes a training-free, interpretable feature truncation-based dimensionality reduction method. The method performs modality-adaptive truncation directly on deep neural network–extracted features (face, fingerprint, and iris) prior to encryption, significantly compressing template size while preserving discriminative information. Its design is natively compatible with homomorphic encryption, enabling secure cross-modal fusion and matching. Experiments on a self-constructed virtual multimodal database demonstrate that the fused templates achieve a 67% size reduction, while the equal error rate (EER) outperforms all single-modal baselines—without compromising recognition accuracy or security; in fact, both are maintained or improved. This work is the first to introduce training-free feature truncation into privacy-preserving multimodal biometrics, achieving an effective balance among computational efficiency, recognition accuracy, and practical deployability.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Biometric recognition is widely used, making the privacy and security of extracted templates a critical concern. Biometric Template Protection schemes, especially those utilizing Homomorphic Encryption, introduce significant computational challenges due to increased workload. Recent advances in deep neural networks have enabled state-of-the-art feature extraction for face, fingerprint, and iris modalities. The ubiquity and affordability of biometric sensors further facilitate multi-modal fusion, which can enhance security by combining features from different modalities. This work investigates the biometric performance of reduced multi-biometric template sizes. Experiments are conducted on an in-house virtual multi-biometric database, derived from DNN-extracted features for face, fingerprint, and iris, using the FRGC, MCYT, and CASIA databases. The evaluated approaches are (i) explainable and straightforward to implement under encryption, (ii) training-free, and (iii) capable of generalization. Dimensionality reduction of feature vectors leads to fewer operations in the Homomorphic Encryption (HE) domain, enabling more efficient encrypted processing while maintaining biometric accuracy and security at a level equivalent to or exceeding single-biometric recognition. Our results demonstrate that, by fusing feature vectors from multiple modalities, template size can be reduced by 67 % with no loss in Equal Error Rate (EER) compared to the best-performing single modality.

Problem

• Reducing computational workload in encrypted biometric systems
• Maintaining accuracy while truncating multi-biometric feature dimensions
• Enabling efficient homomorphic encryption for privacy-preserving fusion

Innovation

• Training-free dimensionality reduction via feature truncation
• Homomorphic Encryption for efficient privacy-preserving processing
• Multi-modal fusion reduces template size by 67%