Filters








27,700 Hits in 5.3 sec

Numerical Composition of Differential Privacy [article]

Sivakanth Gopi, Yin Tat Lee, Lukas Wutschitz
2021 arXiv   pre-print
We give a fast algorithm to optimally compose privacy guarantees of differentially private (DP) algorithms to arbitrary accuracy.  ...  Our method is based on the notion of privacy loss random variables to quantify the privacy loss of DP algorithms.  ...  Numerical Composition of Differential Privacy∗ Sivakanth Gopi1 , Yin Tat Lee2  ... 
arXiv:2106.02848v3 fatcat:hnivivvpgbdv7fx4wgpjt52vem

Numerical Composition of Differential Privacy [article]

Sivakanth Gopi, Yin Tat Lee, Lukas Wutschitz
2021
We give a fast algorithm to optimally compose privacy guarantees of differentially private (DP) algorithms to arbitrary accuracy.  ...  Our method is based on the notion of privacy loss random variables to quantify the privacy loss of DP algorithms.  ...  Numerical composition of privacy curves In this section, we present an efficient and numerically accurate method, ComposePRV (Algorithm 1), for composing privacy guarantees by utilizing the notion of PRVs  ... 
doi:10.48550/arxiv.2106.02848 fatcat:xcucp3v6sfb4tmpd3nw7gww524

Differential identifiability clustering algorithms for big data analysis

Tao Shang, Zheng Zhao, Xujie Ren, Jianwei Liu
2021 Science China Information Sciences  
The definition of ρ-differential identifiability (DI) precisely matches the legal definitions of privacy, which can provide an easy parameterization approach for practitioners so that they can set privacy  ...  DI k-means algorithm uses the usual Laplace mechanism and composition properties for numerical databases, while DI k-prototypes algorithm uses the new exponential mechanism and composition properties for  ...  In fact, there is much non-numerical data required for privacy preservation. (2) Composition of differential identifiability.  ... 
doi:10.1007/s11432-020-2910-1 fatcat:4odlynwnejapjoxgebuhwmrrea

Learning Numeric Optimal Differentially Private Truncated Additive Mechanisms [article]

David M. Sommer, Lukas Abfalterer, Sheila Zingg, Esfandiar Mohammadi
2021 arXiv   pre-print
Differentially private (DP) mechanisms face the challenge of providing accurate results while protecting their inputs: the privacy-utility trade-off.  ...  We introduce a gradient-descent-based tool to learn truncated noise for additive mechanisms with strong utility bounds while simultaneously optimizing for differential privacy under sequential composition  ...  This line shows that a relaxation of differential privacy, called approximate differential privacy (ADP), leads to stronger sequential composition bounds than pure differential privacy.  ... 
arXiv:2107.12957v1 fatcat:yzpktklssvaglp5ushnabnouqm

Improving the Utility of Differential Privacy via Univariate Microaggregation [chapter]

David Sánchez, Josep Domingo-Ferrer, Sergio Martínez
2014 Lecture Notes in Computer Science  
Our proposal relies on univariate microaggregation to reduce the amount of noise needed to satisfy differential privacy.  ...  Differential privacy is a privacy model for anonymization that offers more robust privacy guarantees than previous models, such as k-anonymity and its extensions.  ...  To benefit from such a noise reduction in the case of multivariate data sets, we rely on the following two composition properties of differential privacy. Lemma 2 (Sequential composition [25] ).  ... 
doi:10.1007/978-3-319-11257-2_11 fatcat:f22ifjgrqnfvpfdynm4w6usg24

Sharp Composition Bounds for Gaussian Differential Privacy via Edgeworth Expansion [article]

Qinqing Zheng, Jinshuo Dong, Qi Long, Weijie J. Su
2020 arXiv   pre-print
To address this question, we introduce a family of analytical and sharp privacy bounds under composition using the Edgeworth expansion in the framework of the recently proposed f-differential privacy.  ...  This raises a fundamental question in differential privacy regarding how the overall privacy bound degrades under composition.  ...  Acknowledgments We are grateful to Yicong Jiang for stimulating discussions in the early stages of this work.  ... 
arXiv:2003.04493v2 fatcat:46fecmsfdjhmronvjfbh7vc7ti

Histogram Publication over Numerical Values under Local Differential Privacy

Xu Zheng, Ke Yan, Jingyuan Duan, Wenyi Tang, Ling Tian, Yingjie Wang
2021 Wireless Communications and Mobile Computing  
Local differential privacy has been considered the standard measurement for privacy preservation in distributed data collection.  ...  Therefore, this paper proposes a highly efficient framework for differentially private histogram publication of numerical values in a distributed environment.  ...  Finally, the compositional property of differential privacy can also be merged with the LDP. Theorem 4 (sequential composition [39] ).  ... 
doi:10.1155/2021/8886255 fatcat:fdgkm5ehtvfg5izhhsmmgxjgra

Rényi Differential Privacy

Ilya Mironov
2017 2017 IEEE 30th Computer Security Foundations Symposium (CSF)  
We propose a natural relaxation of differential privacy based on the Renyi divergence.  ...  We demonstrate that the new definition shares many important properties with the standard definition of differential privacy, while additionally allowing tighter analysis of composite heterogeneous mechanisms  ...  comments, and Mark Bun and Thomas Steinke for sharing a draft of [10] .  ... 
doi:10.1109/csf.2017.11 dblp:conf/csfw/Mironov17 fatcat:kyuwkxdxjjehll7chnjmju5n6q

Every Query Counts: Analyzing the Privacy Loss of Exploratory Data Analyses [article]

Saskia Nuñez von Voigt, Mira Pauli, Johanna Reichert, Florian Tschorsch
2020 arXiv   pre-print
In this paper, we quantify the privacy loss for basic statistical functions and highlight the importance of taking it into account when calculating the privacy-loss budget of a machine learning approach  ...  While privacy-preserving machine learning is on the rise, more often than not this initial analysis is not counted towards the privacy budget.  ...  Any mechanism guaranteeing differential privacy is robust under composition [10] .  ... 
arXiv:2008.12282v1 fatcat:nvud55m2dvglnkq7gv3zx7cy6u

Differential Privacy at Risk: Bridging Randomness and Privacy Budget

Ashish Dandekar, Debabrota Basu, Stéphane Bressan
2021 Proceedings on Privacy Enhancing Technologies  
We provide a composition theorem that leverages privacy at risk.  ...  We quantitatively show that composition using the cost optimal privacy at risk provides stronger privacy guarantee than the classical advanced composition.  ...  We thank Pierre Senellart for his help in reviewing derivations of privacy at risk for the Laplace mechanism.  ... 
doi:10.2478/popets-2021-0005 fatcat:vuubbbuzrfbijd2ajfnzrxb2bu

Differential Privacy at Risk: Bridging Randomness and Privacy Budget [article]

Ashish Dandekar, Debabrota Basu, Stephane Bressan
2020 arXiv   pre-print
We quantitatively show that composition using the cost optimal privacy at risk provides stronger privacy guarantee than the classical advanced composition.  ...  The calibration of noise for a privacy-preserving mechanism depends on the sensitivity of the query and the prescribed privacy level.  ...  Acknowledgements We want convey a special thanks to Pierre Senellart at DI,École Normale Supérieure, Paris for his careful reading of our drafts and thoughtful interventions.  ... 
arXiv:2003.00973v2 fatcat:gdt3xf6ho5bgrjvsy5p5x7vssq

Solo: A Lightweight Static Analysis for Differential Privacy [article]

Chike Abuah, David Darais, Joseph P. Near
2021 arXiv   pre-print
of differential privacy, which requires "full" linear types a la Girard.  ...  All current approaches for statically enforcing differential privacy in higher order languages make use of either linear or relational refinement types.  ...  To determine the total privacy cost of a complete program, we need to use the sequential composition property of differential privacy.  ... 
arXiv:2105.01632v2 fatcat:2fd2pfx3ibaullw7z75a5otyqu

Deep Learning with Gaussian Differential Privacy [article]

Zhiqi Bu and Jinshuo Dong and Qi Long and Weijie J. Su
2020 arXiv   pre-print
Leveraging the appealing properties of f-differential privacy in handling composition and subsampling, this paper derives analytically tractable expressions for the privacy guarantees of both stochastic  ...  In this paper, we consider a recently proposed privacy definition termed f-differential privacy [18] for a refined privacy analysis of training neural networks.  ...  We would also like to thank two anonymous referees for their constructive comments that improved the presentation of the paper.  ... 
arXiv:1911.11607v3 fatcat:nd27ura2efdxncz63owe7sc4g4

Proving Differential Privacy via Probabilistic Couplings [article]

Gilles Barthe, Marco Gaboardi, Benjamin Grégoire, Justin Hsu, Pierre-Yves Strub
2017 arXiv   accepted
In this paper, we develop compositional methods for formally verifying differential privacy for algorithms whose analysis goes beyond the composition theorem.  ...  While our paper is presented from a formal verification perspective, we believe that its main insight is of independent interest for the differential privacy community.  ...  We also thank him and Jonathan Ullman for good discussions about challenges and subtleties of the proof of Sparse Vector.  ... 
arXiv:1601.05047v4 fatcat:ebuijcjdzvbalcbqephgryxoyy

APDP: Attack-Proof Personalized Differential Privacy Model for a Smart Home

Yuping Zhang, Youyang Qu, Longxiang Gao, Tom H. Luan, Xi Zheng, Shiping Chen, Yong Xiang
2019 IEEE Access  
INDEX TERMS Smart home, fog computing, differential privacy, personalized privacy.  ...  Then, we apply a personalized differential privacy scheme to provide privacy protection.  ...  Theorem 2 (Attack-Proof): Two privacy levels, 1 and 2 , which represent abbreviated notations of 1  ... 
doi:10.1109/access.2019.2953133 fatcat:w4sw2iucf5dwlbixl6rxy72uii
« Previous Showing results 1 — 15 out of 27,700 results