Filters








413 Hits in 5.2 sec

THE-X: Privacy-Preserving Transformer Inference with Homomorphic Encryption [article]

Tianyu Chen, Hangbo Bao, Shaohan Huang, Li Dong, Binxing Jiao, Daxin Jiang, Haoyi Zhou, Jianxin Li, Furu Wei
2022 arXiv   pre-print
Experiments reveal our proposed THE-X can enable transformer inference on encrypted data for different downstream tasks, all with negligible performance drop but enjoying the theory-guaranteed privacy-preserving  ...  To protect privacy, it is an attractive choice to compute only with ciphertext in homomorphic encryption (HE).  ...  We also thanks the support from Beijing Advanced Innovation Center for Future Blockchain and Privacy Computing.  ... 
arXiv:2206.00216v2 fatcat:mhgrqy2fb5fyhil5u66x6toduq

THE-X: Privacy-Preserving Transformer Inference with Homomorphic Encryption

Tianyu Chen, Hangbo Bao, Shaohan Huang, Li Dong, Binxing Jiao, Daxin Jiang, Haoyi Zhou, Jianxin Li, Furu Wei
2022 Findings of the Association for Computational Linguistics: ACL 2022   unpublished
Experiments reveal our proposed THE-X can enable transformer inference on encrypted data for different downstream tasks, all with negligible performance drop but enjoying the theory-guaranteed privacy-preserving  ...  To protect privacy, it is an attractive choice to compute only with ciphertext in homomorphic encryption (HE).  ...  We also thanks the support from Beijing Advanced Innovation Center for Future Blockchain and Privacy Computing.  ... 
doi:10.18653/v1/2022.findings-acl.277 fatcat:w2m2msqzardynl37jzvjqbbt4q

Blind Faith: Privacy-Preserving Machine Learning using Function Approximation [article]

Tanveer Khan, Alexandros Bakas, Antonis Michalas
2021 arXiv   pre-print
To make our construction compatible with homomorphic encryption, we approximate the activation functions using Chebyshev polynomials.  ...  This allowed us to build a privacy-preserving machine learning model that can classify encrypted images.  ...  The CNN is deployed in a privacy-preserving manner in the CSP. To preserve the users privacy, we use HE. Using an HE scheme allows us to perform computations on encrypted data.  ... 
arXiv:2107.14338v1 fatcat:aceanmkp4vd53egwgr6q34nyqi

Efficient Private Machine Learning by Differentiable Random Transformations [article]

Fei Zheng
2020 arXiv   pre-print
With the increasing demands for privacy protection, many privacy-preserving machine learning systems were proposed in recent years.  ...  However, most of them cannot be put into production due to their slow training and inference speed caused by the heavy cost of homomorphic encryption and secure multiparty computation(MPC) methods.  ...  And let the auxiliary information for adversary is the x and A both are drawn from standard normal distribution.  ... 
arXiv:2008.07758v1 fatcat:hdynkkfqfvdkbkosvgj5pytk7e

Improving Utility of Differentially Private Mechanisms through Cryptography-based Technologies: a Survey [article]

Wen Huang, Shijie Zhou, Tianqing Zhu, Yongjian Liao
2021 arXiv   pre-print
Then, we summarize how to improve utility by combining differentially private mechanisms with homomorphic encryption schemes.  ...  Therefore, privacy preservation has become an urgent problem to be solved. Differential privacy as a strong privacy preservation tool has attracted significant attention.  ...  In particular, for a 1 = 4 the x 1 is 2(4 = 5 2 mod 7). While for a 2 = 1 the x 2 is 0(1 = 5 0 mod 7) and a 3 = 2 the x 3 is 4(2 = 5 4 mod 7).  ... 
arXiv:2011.00976v2 fatcat:fizzcprz55cdxa7bwzyt7rzree

Secure Image Inference using Pairwise Activation Functions

Jonas T. Agyepong, Mostafa Soliman, Yasutaka Wada, Keiji Kimura, Ahmed El-Mahdy
2021 IEEE Access  
INDEX TERMS Exploratory analysis, homomorphic encryption scheme, homomorphic image inference, pairwise functions, polynomial approximation, privacy-preserving machine learning.  ...  Polynomial approximation has for the past few years been used to derive polynomials as an approximation to activation functions for use in image prediction or inference employing homomorphic encryption  ...  and [−10, 10] on the x-axis of the x − y plane and used an interval of 0.5.  ... 
doi:10.1109/access.2021.3106888 fatcat:5jqu6yjkl5hb3l2igp5e6a3nim

Faster CryptoNets: Leveraging Sparsity for Real-World Encrypted Inference [article]

Edward Chou, Josh Beal, Daniel Levy, Serena Yeung, Albert Haque, Li Fei-Fei
2018 arXiv   pre-print
This work increases the viability of deep learning systems that use homomorphic encryption to protect user privacy.  ...  This privacy-preserving feature is attractive for machine learning, but requires significant computational time due to the large overhead of the encryption scheme.  ...  (Top) The x axis denotes the pre-activation value. (Bottom) The x axis denotes the post-activation value. (Both)The y axis denotes a normalized frequency.  ... 
arXiv:1811.09953v1 fatcat:67vqy4vwqvahvd6kqcxuzp23ki

Image Feature Extraction in Encrypted Domain With Privacy-Preserving SIFT

Chao-Yung Hsu, Chun-Shien Lu, Soo-Chang Pei
2012 IEEE Transactions on Image Processing  
As all of the operations in SIFT must be moved to the encrypted domain, we propose a privacy-preserving realization of the SIFT method based on homomorphic encryption.  ...  It is envisioned that secure media applications with privacy preservation will be treated seriously.  ...  not both coinciding with the x− and y− axes.  ... 
doi:10.1109/tip.2012.2204272 pmid:22711774 fatcat:hxe4lcmcxngqzhohiesblb3bjy

Neural Network Training With Homomorphic Encryption [article]

Kentaro Mihara, Ryohei Yamaguchi, Miguel Mitsuishi, Yusuke Maruyama
2020 arXiv   pre-print
Our method relies on homomorphic capability of lattice based encryption scheme.  ...  We introduce a novel method and implementation architecture to train neural networks which preserves the confidentiality of both the model and the data.  ...  Privacy-preserving machine learning methods can be categorized into two main classes: secure multiparty computation, and homomorphic encryption.  ... 
arXiv:2012.13552v1 fatcat:sp3ef7wc3zeqfnbfs7ologujhm

Applying Deep Neural Networks over Homomorphic Encrypted Medical Data

Anamaria Vizitiu, Cosmin Ioan Niƫă, Andrei Puiu, Constantin Suciu, Lucian Mihai Itu
2020 Computational and Mathematical Methods in Medicine  
The findings highlight the potential of the proposed privacy-preserving deep learning methods to outperform existing approaches by providing, within a reasonable amount of time, results equivalent to those  ...  The considered encryption scheme, MORE (Matrix Operation for Randomization or Encryption), enables the computations within a neural network model to be directly performed on floating point data with a  ...  Figure 14 : 14 Figure 14: Confusion matrix of the X-ray coronary angiography view classifier. Table 1 : 1 MORE encryption scheme setup over rational numbers.  ... 
doi:10.1155/2020/3910250 pmid:32351612 pmcid:PMC7171620 fatcat:qbtrygkl3jbnhholjrjdde7rsq

Enhanced Security in Cloud Computing Using Neural Network and Encryption

Muhammad Usman Sana, Zhanli Li, Fawad Javaid, Hannan Bin Liaqat, Muhammad Usman Ali
2021 IEEE Access  
In this paper, to train neural networks using encrypted data we considered the Matrix Operation-based Randomization and Encipherment (MORE) technique, based on Fully Homomorphic Encryption (FHE).  ...  With the fast advancement in cloud computing, progressively more users store their applications and data on the cloud.  ...  Where the number of epoch/iteration is shown on the y-axis and time in sec is shown on the x-axis. B.  ... 
doi:10.1109/access.2021.3122938 fatcat:jpnki543zncbnij37pivanhbvi

Cryptographic Solutions for Genomic Privacy [chapter]

Erman Ayday
2016 Lecture Notes in Computer Science  
With the help of rapidly developing technology, DNA sequencing is becoming less expensive.  ...  In this work, focusing on our existing and ongoing work on genomic privacy, we will first highlight one serious threat for genomic privacy.  ...  For each family member, we reveal 50 randomly picked SNPs (out of 100 SNPs on chromosome 1), starting from the most distant family members, and the x-axis represents the exact sequence of this disclosure  ... 
doi:10.1007/978-3-662-53357-4_22 fatcat:d2ocbtus4jg5hn5csu36nep7im

Privacy-Preserving Federated Neural Network Learning for Disease-Associated Cell Classification [article]

Sinem Sav, Jean-Philippe Bossuat, Juan R. Troncoso-Pastoriza, Manfred Claassen, Jean-Pierre Hubaux
2022 bioRxiv   pre-print
PriCell relies on multiparty homomorphic encryption and enables the collaborative training of encrypted neural networks with multiple healthcare institutions.  ...  We efficiently replicate the training of a published state-of-the-art convolutional neural network architecture in a decentralized and privacy-preserving manner.  ...  The most recent solutions for privacy-preserving federated learning in the N-party setting use differential privacy (DP), secure multiparty computation (SMC), or homomorphic encryption (HE).  ... 
doi:10.1101/2022.01.10.475610 fatcat:uokre4qusnfmbbzaqzrmb6in5e

Privacy-Preserving Locally Weighted Linear Regression over Encrypted Millions of Data

Xiaoxia Dong, Jie Chen, Kai Zhang, Haifeng Qian
2019 IEEE Access  
INDEX TERMS Locally weighted linear regression, privacy-preserving, paillier homomorphic encryption, stochastic gradient descent.  ...  Therefore, we use Paillier homomorphic encryption as the building modular to encrypt data and then apply the stochastic gradient descent in encrypted domain.  ...  homomorphic encryption with more highly efficiency than somewhat homomorphic encryption [18] .  ... 
doi:10.1109/access.2019.2962700 fatcat:pstfwmovp5h4vlceyufbv5laue

Highly Accurate CNN Inference Using Approximate Activation Functions over Homomorphic Encryption [article]

Takumi Ishiyama, Takuya Suzuki, Hayato Yamana
2020 arXiv   pre-print
We implemented CNN inference labeling over homomorphic encryption using the Microsoft's Simple Encrypted Arithmetic Library for the Cheon-Kim-Kim-Song (CKKS) scheme.  ...  In this study, we seek to improve the classification accuracy for inference processing in a convolutional neural network (CNN) while using homomorphic encryption.  ...  Thus, a privacy-preserving machine learning (PPML) capability, which conducts training and inference processing using a machine-learning model while protecting privacy, is being actively pursued.  ... 
arXiv:2009.03727v2 fatcat:cjouuqis2rek3hnzvqbec3o7f4
« Previous Showing results 1 — 15 out of 413 results