ShadowNet: A Secure and Efficient On-device Model Inference System for Convolutional Neural Networks
Google(谷歌);Florida International University(佛罗里达国际大学);Northeastern University(东北大学);University of California, San Diego(加州大学圣地亚哥分校);University of Wisconsin-Madison(威斯康星大学麦迪逊分校)
Zhichuang Sun, Ruimin Sun, Changming Liu, Amrita Roy Chowdhury, Long Lu, Somesh Jha
With the increased usage of AI accelerators on mobile and edge devices, on-device machine learning (ML) is gaining popularity. Thousands of proprietary ML models are being deployed today on billions of untrusted devices. This raises serious security concerns about model privacy. However, protecting model privacy without losing access to the untrusted AI accelerators is a challenging problem. In this paper, we present a novel on-device model inference system, ShadowNet. ShadowNet protects the model privacy with Trusted Execution Environment (TEE) while securely outsourcing the heavy linear layers of the model to the untrusted hardware accelerators. ShadowNet achieves this by transforming the weights of the linear layers before outsourcing them and restoring the results inside the TEE. The non-linear layers are also kept secure inside the TEE. ShadowNet's design ensures efficient transformation of the weights and the subsequent restoration of the results. We build a ShadowNet prototype based on TensorFlow Lite and evaluate it on five popular CNNs, namely, MobileNet, ResNet-44, MiniVGG, ResNet-404, and YOLOv4-tiny. Our evaluation shows that ShadowNet achieves strong security guarantees with reasonable performance, offering a practical solution for secure on-device model inference.
Wuhan University(武汉大学);State Key Laboratory of Cryptology(密码科学技术国家重点实验室);The Chinese University of Hong Kong(香港中文大学);Huazhong University of Science and Technology(华中科技大学);Xi’an Jiaotong University(西安交通大学)
卷积神经网络是一种广泛应用于各种预测任务的机器学习模型,如计算机视觉和医学图像分析。它们强大的预测能力需要大量的计算,这鼓励模型所有者在云平台上托管预测服务。最近的研究集中在查询和结果的隐私,但他们没有提供对模型托管服务器的模型隐私,并可能泄漏部分信息的结果。它们中的一些还需要与查询器的频繁交互或繁重的计算开销,这阻碍了查询器使用预测服务。提出了一种新的外包环境下隐私保护神经网络预测方案,服务器不能学习查询、(中间)结果和模型。类似于SecureML(S&P'17),一个提供模型隐私的代表性工作,我们利用两个具有秘密共享和三元组生成的非合谋服务器来最大限度地减少重量级加密的使用。此外,我们采用异步计算,以提高吞吐量,并设计乱码电路的非多项式激活函数,以保持相同的精度作为底层网络(而不是近似)。我们在MNIST数据集上的实验表明,与SecureML,MiniONN(CCS'17)和EzPC(EuroS& P' 19)相比,我们的方案分别实现了平均122倍,14.63倍和36.69倍的延迟减少。在通信成本方面,我们的方案平均比SecureML高1.09倍,比MiniONN高36.69倍,比EzPC高31.32倍。在CIFAR数据集上,与MiniONN和EzPC相比,我们的方案分别实现了7.14倍和3.48倍的低延迟。在CIFAR数据集上,我们的方案还提供了比MiniONN和EzPC低1…
隐私问题是深度学习中的一个重要问题,即在深度神经网络开发过程中,必须禁止与第三方共享敏感数据。在本文中,我们提出了Morphed Learning(MoLe),这是一种高效安全的深度学习数据传输方案。MoLe有两个主要组件:数据变形和增强卷积(Aug-Conv)层。数据变形允许数据提供者在没有隐私信息的情况下发送变形的数据,而Aug-Conv层帮助深度学习开发人员将其网络应用于变形的数据,而不会带来性能损失。与GAZELLE(USENIX Security 2018)相比,MoLe提供了更强的安全性,同时引入了更低的开销,这是另一种对神经网络没有性能损失的方法。在CIFAR数据集上使用MoLe进行VGG-16网络时,计算开销仅为9%,数据传输开销为5.12%。作为比较,GAZELLE具有10,000次的计算开销和421,000次的数据传输开销。在这种情况下,MoLe和GAZELLE的攻击成功率分别为7.9 x 10^{-90}和2.9 x 10^{-30}。
PrivGNN: High-Performance Secure Inference for Cryptographic Graph Neural Networks
Deakin University(德肯大学);Singapore University of Technology and Design(新加坡科技与设计大学);Fuzhou University(福州市大学);East China Normal University(华东师范大学);Griffith University(格里菲斯大学)
Privacy-Preserving Detection of Rare Disease-Associated Cell Subsets via Secure Multi-Party Computation
Institute for Bioinformatics and Medical Informatics, University of Tübingen(蒂宾根大学生物信息学与医学信息学研究所);Medical Data Privacy and Privacy-Preserving Machine Learning, University of Tübingen(蒂宾根大学医学数据隐私与隐私增强机器学习研究所);University Hospital Aachen(亚琛大学医院);Delft University of Technology(代尔夫特理工大学)
Efficient Encrypted Computation in Convolutional Spiking Neural Networks with TFHE
School of Cyber Science and Engineering, Huazhong University of Science and Technology(华中科技大学网络空间安全学院);School of Mathematics and Statistics, Huazhong University of Science and Technology(华中科技大学数学与统计学院);Center for Mathematical Science, Huazhong University of Science and Technology(华中科技大学数学中心);Steklov-Wuhan Institute for Mathematical Exploration, Huazhong University of Science and Technology(华中科技大学斯捷克洛夫-武汉数学探索研究所);Department of Mathematics and Department of Physics, Great Bay University(大湾区大学数学与物理学院);Guangdong Provincial Key Laboratory of Mathematical and Neural Dynamical Systems(广东省数学与神经动力系统重点实验室)
Hybrid DeepXAI Fusion Network for Interpretable Big Data Analytics and Secure Cloud Decision-Making Systems
Usha Rama College of Engineering and Technology(乌沙·拉玛工程与技术学院);Chaitanya Bharathi Institute of Technology(柴坦亚·巴拉蒂理工学院);VelTech Rangarajan Dr.Sagunthala R&D Institute of Science and Technology(韦尔特·兰加拉詹·德拉·萨贡塔拉研发科学与技术学院);Koneru Lakshmaiah Education Foundation(科内鲁·拉克什马亚教育基金会);Saveetha School of Engineering. Saveetha Institute of Medical and Technical Sciences(萨维塔工程学院。萨维塔医学与科技学院);Vidya Jyothi Institute of Technology (VJIT)(维迪亚·乔蒂理工学院)
FFConv: Fast Factorized Convolutional Neural Network Inference on Encrypted Data
Singapore Management University(新加坡管理大学)
同态加密(HE)允许对加密数据(密文)进行计算而无需首先对其进行解密,从而为云中的隐私保护应用程序提供安全但速度非常慢的卷积神经网络(CNN)推理。为了减少推理延迟,一种方法是将多个消息打包到单个密文中,以便减少密文的数量并支持密文之间的同态乘法累加(HMA)操作的大规模并行性。尽管HECNN推理速度更快,但主流打包方案密集打包(DensePack)和卷积打包(ConvPack)引入了昂贵的旋转开销,这降低了HECNN用于更深和更宽的CNN架构的推理延迟。在本文中,我们提出了一种低秩分解方法FFConv专用于有效的密文包装,以减少旋转开销和HMA操作。FFConv使用低秩分解卷积来近似d x d卷积层,其中具有较少通道的d x d低秩卷积之后是1 x 1卷积以恢复通道。使用DensePack的d x d低秩卷积显著减少了旋转操作,而使用ConvPack的1 x 1卷积的旋转开销接近于零。据我们所知,FFConv是第一个能够同时减少DensePack和ConvPack所产生的旋转开销的工作,而不会向HECNN推理管道中引入额外的特殊块。与现有技术的LoLa和Falcon相比,我们的方法分别将推理延迟降低了88%和21%,并且在MNIST和CIFAR-10上具有相当的准确性。
LinGCN: Structural Linearized Graph Convolutional Network for Homomorphically Encrypted Inference
University of Rochester(罗切斯特大学);Indiana University Bloomington(印第安纳大学伯明顿分校);Northeastern University(东北大学);North Carolina State University(北卡罗来纳州立大学);University of Connecticut(康涅狄格大学);University of Massachusetts Dartmouth(马萨诸塞大学达特茅斯分校)
Localization of Dummy Data Injection Attacks in Power Systems Considering Incomplete Topological Information: A Spatio-Temporal Graph Wavelet Convolutional Neural Network Approach
Northeast Electric Power University(东北电力大学);Jilin Electric Power Research Institute of State Grid(国家电网吉林电力科学研究院);Northeast Electric Power Design and Institute Co., Ltd. of China Power Engineering Consulting Group(中国电力工程顾问集团东北电力设计院有限公司);The Hong Kong University of Science and Technology(香港科技大学)