HERS: Homomorphically Encrypted Representation Search

@article{Engelsma2020HERSHE,
  title={HERS: Homomorphically Encrypted Representation Search},
  author={Joshua J. Engelsma and Anil K. Jain and Vishnu Naresh Boddeti},
  journal={IEEE Transactions on Biometrics, Behavior, and Identity Science},
  year={2020},
  volume={4},
  pages={349-360}
}
We present a method to search for a probe (or query) image representation against a large gallery in the encrypted domain. We require that the probe and gallery images be represented in terms of a fixed-length representation, which is typical for representations obtained from learned networks. Our encryption scheme is agnostic to how the fixed-length representation is obtained and can therefore be applied to any fixed-length representation in any application domain. Our method, dubbed HERS… 

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

A survey of biometric template protection (BTP) methods proposed for securing face “templates” in neural-network-based face recognition systems, and recommends a greater focus on empirical evaluation strategies to provide more concrete insights into the irreversibility and renewability/unlinkability of face BTP methods in practice.

Towards Privacy-Preserving, Real-Time and Lossless Feature Matching

A plug-in module called SecureVector is proposed that protects features by random permutations, 4L-DEC converting and existing homomorphic encryption techniques that achieves real-time and lossless feature matching among sanitized features, along with much higher security levels than current state-of-the-arts.

Feature Fusion Methods for Indexing and Retrieval of Biometric Data: Application to Face Recognition with Privacy Protection

By facilitating a seamless integration of template protection with open-source homomorphic encryption libraries, the proposed method guarantees unlinkability, irreversibility, and renewability of the protected biometric data.

SpoofGAN: Synthetic Fingerprint Spoof Images

This work aims to demonstrate the utility of synthetic (both bona fide and PA style) fingerprints in supplying these algorithms with sufficient data to improve the performance of fingerprint PAD algorithms beyond the capabilities when training on a limited amount of publicly available “real” datasets.

Private Data Preprocessing for Privacy-preserving Federated Learning

  • Ruei-Hau HsuTing-Yun Huang
  • Computer Science
    2022 IEEE 5th International Conference on Knowledge Innovation and Invention (ICKII )
  • 2022
A privacy-preserving federated learning architecture based on homomorphic encryption is designed to protect each participant's data and local model and encrypted query technology is used in this architecture to provide data privacy matching.

Hybrid Protection of Biometric Templates by Combining Homomorphic Encryption and Cancelable Biometrics

A hybrid scheme for the protection of biometric templates is proposed, which combines cancelable biometrics (CB) methods and HE, and can reduce the dimensionality of templates using CB before applying HE, which speeds up the computation over the ciphertexts.

Indexing Protected Deep Face Templates by Frequent Binary Patterns

  • Dailé Osorio RoigC. RathgebHatef Otroshi-ShahrezaC. BuschS. Marcel
  • Computer Science
    2022 IEEE International Joint Conference on Biometrics (IJCB)
  • 2022
This approach represents to the best of the authors' knowledge the first generic indexing scheme that can be applied to arbitrary cancelable face templates (o binary representation) and binning and retrieving cancelable deep face templates based on frequent binary patterns.

Efficient Biometric Identification on the Cloud With Privacy Preservation Guarantee

Two new privacy-preserving biometric identification outsourcing protocols are proposed, one mainly utilizes the efficient Householder transformation and permutation technique and the other initializes a novel random split technique and combines it with the invertible linear transformation to achieve a higher security requirement under the known-plaintext attack model.

Minutiae-Guided Fingerprint Embeddings via Vision Transformers

This work proposes the first use of a Vision Transformer (ViT) to learn a discriminative minutiae matching embedding and shows that by fusing embeddings learned by CNNs and ViTs the authors can reach near parity with a commercial state-of-the-art (SOTA) matcher.

HEFT: Homomorphically Encrypted Fusion of Biometric Templates

This paper proposes a non-interactive end-to-end solution for secure fusion and matching of biometric templates using fully homomorphic encryption (FHE) and introduces an FHE-aware algorithm for learning the linear projection matrix to mitigate errors induced by approximate normalization.

References

SHOWING 1-10 OF 87 REFERENCES

Secure Face Matching Using Fully Homomorphic Encryption

The practicality of using a fully homomorphic encryption based framework to secure a database of face templates, designed to preserve the privacy of users and prevent information leakage from the templates, while maintaining their utility through template matching directly in the encrypted domain is explored.

The MegaFace Benchmark: 1 Million Faces for Recognition at Scale

The MegaFace dataset is assembled, both for identification and verification performance, and performance with respect to pose and a persons age is evaluated, as a function of training data size (#photos and #people).

Learning a Fixed-Length Fingerprint Representation

DeepPrint incorporates fingerprint domain knowledge, including alignment and minutiae detection, into the deep network architecture to maximize the discriminative power of its representation, which is the most compact and discrim inative fixed-length fingerprint representation reported in the academic literature.

On the Intrinsic Dimensionality of Image Representations

A deep neural network based non-linear mapping is developed, dubbed DeepMDS, that transforms the ambient representation to the minimal intrinsic space, and the veracity of the mapping through image matching in the intrinsic space is validated.

Homomorphic Encryption for Arithmetic of Approximate Numbers

We suggest a method to construct a homomorphic encryption scheme for approximate arithmetic. It supports an approximate addition and multiplication of encrypted messages, together with a new

Image Reconstruction from Bag-of-Visual-Words

This work uses a large-scale image database to estimate the spatial arrangement of local descriptors and proposes a heuristic but efficient method to optimize it, which can reconstruct the original images, although the image features lack spatial information and include quantization errors.

IronMask: Modular Architecture for Protecting Deep Face Template

This paper presents a modular architecture for face template protection, called IronMask, that can be combined with any face recognition system using angular distance metric, and circumvent the need for binarization by proposing a new real-valued error-correcting-code that is compatible withreal-valued templates and can therefore, minimize performance degradation.

Efficient Encrypted Inference on Ensembles of Decision Trees

This work proposes a framework to transfer knowledge extracted by complex decision tree ensembles to shallow neural networks (referred to as DTNets) that are highly conducive to encrypted inference and minimizes the accuracy loss by searching for the best DTNet architecture that operates within the given depth constraints.

Efficient CNN Building Blocks for Encrypted Data

This work considers a Machine Learning as a Service (MLaaS) scenario where both input data and model parameters are secured using FHE, and shows that operational parameters of the chosen FHE scheme such as the degree of the cyclotomic polynomial, depth limitations of the underlying leveled HE scheme, and the computational precision parameters have a major impact on the design of the machine learning model.

Efficiency Analysis of Post-quantum-secure Face Template Protection Schemes based on Homomorphic Encryption

It is shown that a face verification in the encrypted domain requires only 50 ms transaction time and a template size of 5.5 KB, and homomorphic encryption allows to compute the distance between two protected templates in theencrypted domain without a degradation of biometric performance with respect to the corresponding system.
...