Biometric Template Protection Through Adversarial Learning

Jami, Santosh Kumar; Chalamala, Srinivasa Rao; Jindal, Arun Kumar

doi:10.1109/icce.2019.8661905

Cited by 15 publications

(28 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These methods learn the protected template from the input face image (with face embeddings being extracted, in some format, during the process). In [13]- [16], a random code is pre-defined for each subject during enrollment, then the neural network is trained to map different samples of the same subject's face to their (same) corresponding code. A cryptographic hash of the random code represents the protected face template.…”

Section: Face Embedding Protection Methodsmentioning

confidence: 99%

“…As stated earlier, however, an overdetermined system of equations may result in an inconsistent solution set, thereby actually confusing the solver as to which solution is the correct one. This may explain the sudden drop in 16. The k values for overlaps 0 to 4 are indicated in Table 1.…”

Section: Multiple Polyprotected Templatesmentioning

confidence: 98%

“…So, the first step was to set up a system of k × p equations for each overlap amount 16 , and for each value of p in the range [1,10]. For example, for overlap = 0, to simulate an attacker with access to the minimum of only one 26-dimensional P , we set up a system of 26 equations in 128 unknowns (embedding elements), same as for the irreversibility analysis in Section 4.3.2.…”

Section: Multiple Polyprotected Templatesmentioning

confidence: 99%

See 2 more Smart Citations

Towards Protecting Face Embeddings in Mobile Face Verification Scenarios

Hahn,

Marcel

2021

Preprint

View full text Add to dashboard Cite

This paper proposes PolyProtect, a method for protecting the sensitive face embeddings that are used to represent people's faces in neural-network-based face verification systems. PolyProtect transforms a face embedding to a more secure template, using a mapping based on multivariate polynomials parameterised by user-specific coefficients and exponents. In this work, PolyProtect is evaluated on two open-source face verification systems in a mobile application context, under the toughest threat model that assumes a fully-informed attacker with complete knowledge of the system and all its parameters. Results indicate that PolyProtect can be tuned to achieve a satisfactory trade-off between the recognition accuracy of the PolyProtected face verification system and the irreversibility of the PolyProtected templates. Furthermore, PolyProtected templates are shown to be effectively unlinkable, especially if the user-specific parameters employed in the PolyProtect mapping are selected in a non-naive manner. The evaluation is conducted using practical methodologies with tangible results, to present realistic insight into the method's robustness as a face embedding protection scheme in practice. The code to fully reproduce this work is available at: https://gitlab.idiap.ch/bob/bob.paper.polyprotect 2021.

show abstract

Section: Face Embedding Protection Methodsmentioning

confidence: 99%

Section: Multiple Polyprotected Templatesmentioning

confidence: 98%

Section: Multiple Polyprotected Templatesmentioning

confidence: 99%

See 1 more Smart Citation

Towards Protecting Face Embeddings in Mobile Face Verification Scenarios

Hahn,

Marcel

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Examples of NN-learned face BTP methods that pre-define a representative binary code (and cryptographically hash it to produce the protected template), include [37]- [42]. Of these, [37]- [39] adopt an image-level approach, whereby the NN is trained in an end-to-end manner to map the input face images to their corresponding, pre-defined codes.…”

Section: Feature-levelmentioning

confidence: 99%

“…Of these, [37]- [39] adopt an image-level approach, whereby the NN is trained in an end-to-end manner to map the input face images to their corresponding, pre-defined codes. Alternatively, [40]- [42] adopt a feature-level approach, where the NN training starts from the extracted face features as opposed to the raw images. Overall, most of these methods build upon the approach proposed in [37], [38], which involves assigning a random maximum-entropy binary code to every user of the face recognition system, then training a Convolutional Neural Network (CNN) to map each user's face image to their corresponding code.…”

Section: Feature-levelmentioning

confidence: 99%

Biometric Template Protection for Neural-Network-based Face Recognition Systems: A Survey of Methods and Evaluation Techniques

Hahn¹,

Marcel²

2021

Preprint

View full text Add to dashboard Cite

As automated face recognition applications tend towards ubiquity, there is a growing need to secure the sensitive face data used within these systems. This paper presents a survey of biometric template protection (BTP) methods proposed for securing face "templates" (images or representative features) in neural-network-based face recognition systems. The BTP methods are categorised into two types: Non-NN and NN-learned. Non-NN methods use a neural network (NN) as a feature extractor, but the BTP part is based on a non-NN algorithm applied at either image-level or feature-level. In contrast, NN-learned methods specifically employ a NN to learn a protected template from the unprotected face image or features. We present examples of Non-NN and NN-learned face BTP methods from the literature, along with a discussion of the two categories' comparative strengths and weaknesses. We also investigate the techniques used to evaluate these BTP methods, in terms of the three most common BTP criteria: "recognition accuracy", "irreversibility", and "renewability/unlinkability". As expected, the recognition accuracy of protected face recognition systems is generally evaluated using the same (empirical) techniques employed for evaluating standard (unprotected) biometric systems. On the contrary, most irreversibility and renewability/unlinkability evaluations are found to be based on theoretical assumptions/estimates or verbal implications, with a lack of empirical validation in a practical face recognition context. We recommend, therefore, a greater focus on empirical evaluation strategies, to provide more concrete insights into the irreversibility and renewability/unlinkability of face BTP methods in practice. Additionally, an exploration of the reproducibility of the studied BTP works, in terms of the public availability of their implementation code and evaluation datasets/procedures, suggests that it would currently be difficult for the BTP community to faithfully replicate (and thus validate) most of the reported findings. So, we advocate for a push towards reproducibility, in the hope of furthering our understanding of the face BTP research field.

show abstract