HyDia: FHE-based Facial Matching with Hybrid Approximations and Diagonalization
Published in Privacy Enhancing Technologies Symposium (PETS) 2025, 2025
Recommended citation: Sam Martin, Nirajan Koirala, Helena Berens, Thomas Rozgonyi, Micah Brody, Taeho Jung. (2025). 'HyDia: FHE-based Facial Matching with Hybrid Approximations and Diagonalization.' Proceedings on Privacy Enhancing Technologies, 2025(3). https://petsymposium.org/popets/2025/popets-2025-0146.php
HyDia is a privacy-preserving facial matching system utilizing fully homomorphic encryption (FHE). The framework introduces hybrid approximation techniques and diagonalization strategies to significantly reduce computational complexity while maintaining accuracy in encrypted biometric comparisons. By combining efficient ciphertext transformations with approximation-aware matching, HyDia achieves practical performance for large-scale deployments without sacrificing security.
Recommended citation: Sam Martin, Nirajan Koirala, Helena Berens, Thomas Rozgonyi, Micah Brody, Taeho Jung. (2025). “HyDia: FHE-based Facial Matching with Hybrid Approximations and Diagonalization.” Proceedings on Privacy Enhancing Technologies, 2025(3).