Journal Article
Training Differentially Private Models with Secure Multiparty Computation
Sikha Pentyala,Davis Railsback,Ricardo Maia,Rafael Dowsley,David Melanson,Anderson C. A. Nascimento,M. De Cock +6 more
TL;DR: The problem of learning a machine learning model from training data that originates at multiple data owners, while providing formal privacy guarantees regarding the protection of each owner’s data is addressed.
read more
Abstract: We address the problem of learning a machine learning model from training data that originates at multiple data owners while providing formal privacy guarantees regarding the protection of each owner's data. Existing solutions based on Differential Privacy (DP) achieve this at the cost of a drop in accuracy. Solutions based on Secure Multiparty Computation (MPC) do not incur such accuracy loss but leak information when the trained model is made publicly available. We propose an MPC solution for training DP models. Our solution relies on an MPC protocol for model training, and an MPC protocol for perturbing the trained model coefficients with Laplace noise in a privacy-preserving manner. The resulting MPC+DP approach achieves higher accuracy than a pure DP approach while providing the same formal privacy guarantees. Our work obtained first place in the iDASH2021 Track III competition on confidential computing for secure genome analysis.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Figures
Table 4: Accuracy of models trained with πLR +πDP for different values of ϵ 
Table 5: Accuracy results obtained with 5-fold CV for ϵ-DP with ϵ = 1 and 2 data owners 
Table 1: Results for ϵ-DP with ϵ = 3 and data from two data owners, as provided by the iDASH2021 competition organizers 
Table 2: 5-fold CV accuracy results for varying number of data owners for ϵ-DP with ϵ = 1. 
Table 7: Accuracy averaged over 5-fold CV with Λ = 1 
Table 8: Runtimes of πLR +πDP for different number r of computing parties
Citations
Semi-Private Computation of Data Similarity with Controlled Leakage
TL;DR: This work devise multiparty computation-protocols to compute similarity of two data sets based on correlation, while offering controllable privacy guarantees and develop methods to compute exact and approximate correlation, respectively, with controlled information leakage.
3
DP-BREM: Differentially-Private and Byzantine-Robust Federated Learning with Client Momentum
Xiao Gu,Ming Li,Lishuang Xiong +2 more
TL;DR: In this article , the authors focus on simultaneously achieving differential privacy and Byzantine robustness for cross-silo federated learning (FL) based on the idea of learning from history.
3
Semi-Private Computation of Data Similarity With Applications to Data Valuation and Pricing
TL;DR: This work devise multiparty computation-protocols to compute similarity of two data sets based on correlation, while offering controllable privacy guarantees and develop methods to compute exact and approximate correlation, respectively, with controlled information leakage.
3
A Decentralized Information Marketplace Preserving Input and Output Privacy
Steven Golob,Sikha Pentyala,Rafael Dowsley,Bernardo David,Mario Larangeira,M. De Cock,Anderson C. A. Nascimento,A. Decentralized +7 more
- 18 Jun 2023
TL;DR: This work proposes a decentralized information marketplace where data held by data providers can be made available for computation to data consumers and enables this privacy-preserving data exchange through a novel and carefully designed combination of a blockchain that supports smart contracts and two privacy-enhancing technologies.
2
Journal Article
Interoperable Private Attribution: A Distributed Attribution and Aggregation Protocol
Benjamin M. Case,Richa Jain,Daniel Masny,Ben Savage,Erik Taubeneck,E. Thomson +5 more
TL;DR: Interoperable Private Attribution (IPA) as mentioned in this paper is a protocol that uses the combination of multi-party computation and differential privacy that enables the processing of peoples' data such that only aggregate measurements are revealed, strictly limiting the information leakage about individual people.
2
References
•Book
The Algorithmic Foundations of Differential Privacy
Cynthia Dwork,Aaron Roth +1 more
- 11 Aug 2014
TL;DR: The preponderance of this monograph is devoted to fundamental techniques for achieving differential privacy, and application of these techniques in creative combinations, using the query-release problem as an ongoing example.
Advances and open problems in federated learning
Peter Kairouz,H. Brendan McMahan,Brendan Avent,Aurélien Bellet,Mehdi Bennis,Arjun Nitin Bhagoji,Kallista Bonawitz,Zachary Charles,Graham Cormode,Rachel Cummings,Rafael G. L. D'Oliveira,Hubert Eichner,Salim El Rouayheb,David Evans,Josh Gardner,Zachary Garrett,Adrià Gascón,Badih Ghazi,Phillip B. Gibbons,Marco Gruteser,Zaid Harchaoui,Chaoyang He,Lie He,Zhouyuan Huo,Ben Hutchinson,Justin Hsu,Martin Jaggi,Tara Javidi,Gauri Joshi,Mikhail Khodak,Jakub Konecní,Aleksandra Korolova,Farinaz Koushanfar,Sanmi Koyejo,Tancrède Lepoint,Yang Liu,Prateek Mittal,Mehryar Mohri,Richard Nock,Ayfer Ozgur,Rasmus Pagh,Hang Qi,Daniel Ramage,Ramesh Raskar,Mariana Raykova,Dawn Song,Weikang Song,Sebastian U. Stich,Ziteng Sun,Ananda Theertha Suresh,Florian Tramèr,Praneeth Vepakomma,Jianyu Wang,Li Xiong,Zheng Xu,Qiang Yang,Felix X. Yu,Han Yu,Sen Zhao +58 more
- 23 Jun 2021
TL;DR: In this article, the authors describe the state-of-the-art in the field of federated learning from the perspective of distributed optimization, cryptography, security, differential privacy, fairness, compressed sensing, systems, information theory, and statistics.
Deep Learning with Differential Privacy
TL;DR: This work develops new algorithmic techniques for learning and a refined analysis of privacy costs within the framework of differential privacy, and demonstrates that deep neural networks can be trained with non-convex objectives, under a modest privacy budget, and at a manageable cost in software complexity, training efficiency, and model quality.
4.3K
Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures
Matt Fredrikson,Somesh Jha,Thomas Ristenpart +2 more
- 12 Oct 2015
TL;DR: A new class of model inversion attack is developed that exploits confidence values revealed along with predictions and is able to estimate whether a respondent in a lifestyle survey admitted to cheating on their significant other and recover recognizable images of people's faces given only their name.
3.3K