APAS: Application-Specific Accelerators for RLWE-Based Homomorphic Linear Transformations

APAS: Application-Specific Accelerators for RLWE-Based Homomorphic Linear Transformations
复制标题

APAS:基于 RLWE 的同态线性变换的特定应用加速器

DOI:
10.1109/tifs.2021.3114032
复制
发表时间:
2021
影响因子:
6.8
通讯作者:
Sato Takashi
Sato Takashi
中科院分区:
计算机科学1区
文献类型:
--
作者:
Bian Song;Kundi Dur E. Shahwar;Hirozawa Kazuma;Liu Weiqiang;Sato Takashi

文献摘要

参考文献

相似文献

近年来,基于同态加密的多方安全计算方案在机器学习领域的应用引起了各研究领域的广泛关注。以往的研究表明,安全协议采用打包加法同态加密(PAHE)方案的基础上环学习错误(RLWE)问题表现出显着的实际优点,特别是在实现高效的安全推理机器学习服务应用程序的前景。在这项工作中,我们介绍了一种新的技术进行同态线性变换(HLT)的PAHE密文。使用所提出的HLT技术,可以在不使用现有工作中提出的数论变换和旋转相加算法的情况下执行同态卷积和内积。为了最大限度地提高HLT技术的效率,我们提出了APAS,一个硬件-软件协同设计框架,由近似算术单元组成的硬件加速HLT。在实验中,我们使用实际的神经网络架构作为基准,以表明APAS可以提高同态卷积的计算和通信效率分别为<inline-formula><tex-math notation="LaTeX">$8\times $</tex-math></inline-formula>和<inline-formula><tex-math notation="LaTeX">$3\times $</tex-math></inline-formula>,与现有方法的ASIC实现相比,能量减少高达<inline-formula><tex-math notation="LaTeX">$26\times $</tex-math></inline-formula>。
Recently, the application of multi-party secure computing schemes based on homomorphic encryption in the field of machine learning attracts attentions across the research fields. Previous studies have demonstrated that secure protocols adopting packed additive homomorphic encryption (PAHE) schemes based on the ring learning with errors (RLWE) problem exhibit significant practical merits, and are particularly promising in enabling efficient secure inference in machine-learning-as-a-service applications. In this work, we introduce a new technique for performing homomorphic linear transformation (HLT) over PAHE ciphertexts. Using the proposed HLT technique, homomorphic convolutions and inner products can be executed without the use of number theoretic transform and the rotate-and-add algorithms that were proposed in existing works. To maximize the efficiency of the HLT technique, we propose APAS, a hardware-software co-design framework consisting of approximate arithmetic units for the hardware acceleration of HLT. In the experiments, we use actual neural network architectures as benchmarks to show that APAS can improve the computational and communicational efficiency of homomorphic convolution by <inline-formula> <tex-math notation="LaTeX">$8\times $ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$3\times $ </tex-math></inline-formula>, respectively, with an energy reduction of up to <inline-formula> <tex-math notation="LaTeX">$26\times $ </tex-math></inline-formula> as compared to the ASIC implementations of existing methods.
DOI: 10.1109/tc.2017.2686385
发表时间: 2017-09
影响因子: 3.7
作者:
S. Roy;F. Vercauteren;Jo Vliegen;Ingrid Verbauwhede
通讯作者: S. Roy;F. Vercauteren;Jo Vliegen;Ingrid Verbauwhede
DOI: 10.1007/978-3-642-13013-7_25
发表时间: 2010-05
期刊: IACR Cryptol. ePrint Arch.
影响因子: --
作者:
N. Smart;F. Vercauteren
通讯作者: N. Smart;F. Vercauteren
DOI: 10.1007/3-540-47721-7_24
发表时间: 1986
期刊: --
影响因子: --
作者:
P. Barrett
通讯作者: P. Barrett
DOI: 10.23919/date.2019.8715110
发表时间: 2019-03
期刊: 2019 Design, Automation & Test in Europe Conference & Exhibition (DATE)
影响因子: --
作者:
S. Bian;Masayuki Hiromoto;Takashi Sato
通讯作者: S. Bian;Masayuki Hiromoto;Takashi Sato
DOI: --
发表时间: 2018-11
期刊: ArXiv
影响因子: --
作者:
Edward Chou;Josh Beal;Daniel Levy;Serena Yeung;Albert Haque;Li Fei-Fei-Li-Fei-Fei-48004138
通讯作者: Edward Chou;Josh Beal;Daniel Levy;Serena Yeung;Albert Haque;Li Fei-Fei-Li-Fei-Fei-48004138