CHIMERA: Combining Ring-LWE-based Fully Homomorphic Encryption Schemes

Boura, Christina; Gama, Nicolas; Georgieva, Mariya; Jetchev, Dimitar

doi:10.1515/jmc-2019-0026

Cited by 82 publications

(50 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Concerning arithmetic types, we highlight the need for boolean and floating-point arithmetic types for linear and non-linear activations (respectively). In this line, hybrid approaches have shown a strong impact, and solutions such as CHIMERA [97] open new paths for optimization. While authors have used integer and boolean arithmetic for discretized and binary neural networks, these are rare and difficult to adapt to modern DL solutions for complex problems.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

SoK: Privacy-Preserving Computation Techniques for Deep Learning

Cabrero-Holgueras

Pastrana

2021

Proceedings on Privacy Enhancing Technologies

View full text Add to dashboard Cite

Deep Learning (DL) is a powerful solution for complex problems in many disciplines such as finance, medical research, or social sciences. Due to the high computational cost of DL algorithms, data scientists often rely upon Machine Learning as a Service (MLaaS) to outsource the computation onto third-party servers. However, outsourcing the computation raises privacy concerns when dealing with sensitive information, e.g., health or financial records. Also, privacy regulations like the European GDPR limit the collection, distribution, and use of such sensitive data. Recent advances in privacy-preserving computation techniques (i.e., Homomorphic Encryption and Secure Multiparty Computation) have enabled DL training and inference over protected data. However, these techniques are still immature and difficult to deploy in practical scenarios. In this work, we review the evolution of the adaptation of privacy-preserving computation techniques onto DL, to understand the gap between research proposals and practical applications. We highlight the relative advantages and disadvantages, considering aspects such as efficiency shortcomings, reproducibility issues due to the lack of standard tools and programming interfaces, or lack of integration with DL frameworks commonly used by the data science community.

show abstract

Section: Discussionmentioning

confidence: 99%

“…For HE, CHIMERA [97] presents a framework that allows for switching between three main HE schemes without the need for decryption. Concretely, it proposes using BFV [58], HEAAN (CKKS [62]) and TFHE [61] for integer, floating-point and boolean arithmetic respectively.…”

Section: Hybrid Techniquesmentioning

confidence: 99%

SoK: Privacy-Preserving Computation Techniques for Deep Learning

Cabrero-Holgueras

Pastrana

2021

Proceedings on Privacy Enhancing Technologies

View full text Add to dashboard Cite

show abstract

“…Both [CDKS20] and [BGGJ20] support conversion of ciphertexts from a lightweight LWE encryption to an RLWE-based HE encryption. The former [CDKS20] has low ciphertext-to-message size expansion and supports packing.…”

Section: Related Workmentioning

confidence: 99%

SEAL-Embedded: A Homomorphic Encryption Library for the Internet of Things

Natarajan¹,

Dai

2021

TCHES

View full text Add to dashboard Cite

The growth of the Internet of Things (IoT) has led to concerns over the lack of security and privacy guarantees afforded by IoT systems. Homomorphic encryption (HE) is a promising privacy-preserving solution to allow devices to securely share data with a cloud backend; however, its high memory consumption and computational overhead have limited its use on resource-constrained embedded devices. To address this problem, we present SEAL-Embedded, the first HE library targeted for embedded devices, featuring the CKKS approximate homomorphic encryption scheme. SEAL-Embedded employs several computational and algorithmic optimizations along with a detailed memory re-use scheme to achieve memory efficient, high performance CKKS encoding and encryption on embedded devices without any sacrifice of security. We additionally provide an “adapter” server module to convert data encrypted by SEAL-Embedded to be compatible with the Microsoft SEAL library for homomorphic encryption, enabling an end-to-end solution for building privacy-preserving applications. For a polynomial ring degree of 4096, using RNS primes of 30 or fewer bits, our library can be configured to use between 64–137 KB of RAM and 1–264 KB of flash data, depending on developer-selected configurations and tradeoffs. Using these parameters, we evaluate SEAL-Embedded on two different IoT platforms with high performance, memory efficient, and balanced configurations of the library for asymmetric and symmetric encryption. With 136 KB of RAM, SEAL-Embedded can perform asymmetric encryption of 2048 single-precision numbers in 77 ms on the Azure Sphere Cortex-A7 and 737 ms on the Nordic nRF52840 Cortex-M4.

show abstract

“…Our methods make use of the homomorphic encryption (HE) formalism (Gentry, 2009) that provides mathematically provable, and potentially one of the strongest security guarantees for protecting genotype data while imputation is performed in an untrusted semi-honest environment. To include a comprehensive set of approaches, we focus on three state-of-the-art HE cryptosystems, namely, Brakerski/Fan-Vercauteren (BFV) (Brakerski, 2012;Fan and Vercauteren, 2012), Cheon-Kim-Kim-Song (CKKS) (Cheon et al, 2017), and the fully homomorphic encryption over the torus (TFHE) (Chillotti et al, 2020;Boura et al, 2018). In our HE-based framework, genotype data are encrypted by the data owner before outsourcing the data.…”

Section: Introductionmentioning

confidence: 99%