Locality preserving binary face representations using auto‐encoders

Abstract

Crypto‐biometric schemes, such as fuzzy commitment, require binary sources. A novel approach to binarising biometric data using Deep Neural Networks applied to facial biometric data is introduced. The binary representations are evaluated on the MOBIO and the Labelled Faces in the Wild databases, where their biometric recognition performance and entropy are measured. The proposed binary embeddings give a state‐of‐the‐art performance on both databases with almost negligible degradation compared to the baseline. The representations' length can be controlled. Using a pretrained convolutional neural network and training the model on a cleaned version of the MS‐celeb‐1M database, binary representations of length 4096 bits and 3300 bits of entropy are obtained. The extracted representations have high entropy and are long enough to be used in crypto‐biometric systems, such as fuzzy commitment. Furthermore, the proposed approach is data‐driven and constitutes a locality preserving hashing that can be leveraged for data clustering and similarity searches. As a use case of the binary representations, a cancellable system is created based on the binary embeddings using a shuffling transformation with a randomisation key as a second factor.

Revocable Crypto-Biometric Key Regeneration Using Face Biometrics, Fuzzy Commitment, and a Cohort Bit Selection Process

“…The system comprises two steps: (i) Extracting binary template from face biometrics and (ii) regenerating the crypto-biometric key using a fuzzy commitment scheme combining the face binary template, which is protected using shuffling protection scheme, and error-correcting codes. The first step is described in our paper [16] and summarized in Section 3.2. The challenge is to obtain long binary representations with high entropy and high biometric performance.…”

“…In Ref. [16], we use a data-driven template binarization method using deep neural networks (DNNs), which does not degrade the performance of the baseline system. Furthermore, we seek to obtain long binary representations with high entropy to be used in crypto-biometric key regeneration schemes.…”

“…The binarization method is detailed in Ref. [16]. Using a pre-trained convolutional neural network (CNN) and training the model on a cleaned version [19] of the MS-celeb-1 M database, we obtain binary representations of length 4096 bits and 3300 bits of entropy.…”

“…We have shown in Ref. [16] that the shuffling scheme is fully unlinkable and irreversible, provided that the user secret second factor is not compromised. The revocability is also provided by the means of the shuffling key; we can re-enroll the user using a different shuffling key when the enrollment is compromised.…”

A Comprehensive Survey on Backdoor Attacks and Their Defenses in Face Recognition Systems

“…Here, we note that an optional feature binarization module [46] may be added after the extractor. Binarization converts a real-valued feature vector into a binary format that enables specific security features such as revocability and nonreusability of the biometric features thanks to cryptographic schemes like Fuzzy Commitment [46], [47].…”