13,664 Hits in 2.5 sec

Differentially Private Matrix Completion Revisited [article]

Prateek Jain, Om Thakkar, Abhradeep Thakurta
2018 arXiv   pre-print
We provide the first provably joint differentially private algorithm with formal utility guarantees for the problem of user-level privacy-preserving collaborative filtering.  ...  Along the way, we provide an optimal differentially private algorithm for singular vector computation, based on the celebrated Oja's method, that provides significant savings in terms of space and time  ...  B Private matrix completion via singular value decomposition (SVD) In this section, we study a simple SVD-based algorithm for differentially private matrix completion.  ... 
arXiv:1712.09765v2 fatcat:t7xc6nurxbcxdl357c5ytq3wtq

A Graph Symmetrisation Bound on Channel Information Leakage under Blowfish Privacy [article]

Tobias Edwards, Benjamin I. P. Rubinstein, Zuhe Zhang, Sanming Zhou
2021 arXiv   pre-print
differential privacy in computer science.  ...  To bound the min-entropy leakage of Blowfish-private mechanisms we organise our analysis over symmetrical partitions corresponding to orbits of graph automorphism groups.  ...  Blowfish privacy in the complete secret graph case coincides with differential privacy.  ... 
arXiv:2007.05975v3 fatcat:ufhhiwecnbc5xlqqklaryws34q

Enabling Privacy-Preserving GWAS in Heterogeneous Human Populations [article]

Sean Simmons, Cenk Sahinalp, Bonnie Berger
2016 arXiv   pre-print
Unfortunately, previous approaches for performing differentially private GWAS are based on rather simple statistics that have some major limitations.  ...  We test our differentially private statistics, PrivSTRAT and PrivLMM, on both simulated and real GWAS datasets and find that they are able to protect privacy while returning meaningful GWAS results.  ...  Our Contribution Previous work on differentially private GWAS have completely ignored the problem of population stratification, greatly limiting its applicability in the real world [17] .  ... 
arXiv:1604.04484v1 fatcat:lni2olvy2rgktjm2kgopp5gfgi

Not Just Cloud Privacy: Protecting Client Privacy in Teacher-Student Learning [article]

Lichao Sun, Ji Wang, Philip S. Yu, Lifang He
2020 arXiv   pre-print
In this work, we re-design the privacy-preserving "teacher-student" model consisting of adopting both private arbitrary masking and local differential privacy, which protects the sensitive information  ...  One recent popular approach to study these concerns is using the differential privacy via a "teacher-student" model, wherein the teacher provides the student with useful, but noisy, information, hopefully  ...  Local Differential Privacy In this section, we revisit the definition of local differential privacy, which is a concept of privacy tailored to the privacy-preserving data analysis.  ... 
arXiv:1910.08038v2 fatcat:6xu7fcftjncgxleadbnkvhbvcu

Differentially Private Fractional Frequency Moments Estimation with Polylogarithmic Space [article]

Lun Wang, Iosif Pinelis, Dawn Song
2021 arXiv   pre-print
We prove that 𝔽_p sketch, a well-celebrated streaming algorithm for frequency moments estimation, is differentially private as is when p∈(0, 1].  ...  𝔽_p sketch uses only polylogarithmic space, exponentially better than existing DP baselines and only worse than the optimal non-private baseline by a logarithmic factor.  ...  Revisiting F p Sketch For completeness, we revisit the well-celebrated F p sketch by [21] (also known as stable projection or compressed counting).  ... 
arXiv:2105.12363v4 fatcat:ybzcqrzxybdyrejhfyrdnv3zi4

A Two-Stage Architecture for Differentially Private Kalman Filtering and LQG Control [article]

Kwassi H. Degue, Jerome Le Ny
2020 arXiv   pre-print
Numerical simulations illustrate the performance improvements over differentially private algorithms without first-stage signal aggregation.  ...  This article revisits the Kalman filtering and Linear Quadratic Gaussian (LQG) control problems, subject to privacy constraints.  ...  private signal ỹi , and hence does not is (, δ)-differentially private.  ... 
arXiv:1707.08919v2 fatcat:5selxv4e5nbvrdztws7we4ps5m

Preserving Differential Privacy in Degree-Correlation based Graph Generation

Yue Wang, Xintao Wu
2013 Transactions on Data Privacy  
The idea is to enforce differential privacy on graph model parameters learned from the original network and then generate the graphs for releasing using the graph model with the private parameters.  ...  Empirical evaluations show the developed private dK-graph generation models significantly outperform the approach based on the stochastic Kronecker generation model.  ...  Empirical evaluations show the effectiveness of our proposed private dK-graph models. Background Differential Privacy We revisit the formal definition and the mechanism of differential privacy.  ... 
pmid:24723987 pmcid:PMC3979555 fatcat:jvqlnv7ftvbp5fvhsivjbezwwi

Privacy-preserving Constrained Spectral Clustering Algorithm for Large-scale Data Sets

Wenfen Liu, Ji Li, Jianghong Wei, Mao Ye, Xuexian Hu
2019 IET Information Security  
Specifically, by combining the well-studied constrained spectral clustering with the Wishart mechanism in a novel way, the authors propose a differentially private constrained spectral clustering (DP-CSC  ...  The DP-CSC algorithm is proved to capture asymptotic property and achieves ϵ-differential privacy.  ...  [46] revisited the variation of this method, proposed a noise algorithm for covariance matrix based on the Guassian mechanism and proved that it satisfies (ϵ, δ)-differential privacy.  ... 
doi:10.1049/iet-ifs.2019.0255 fatcat:qpk72f4shzb45k6aomtbrxcbwm

A Differential Privacy Mechanism Design Under Matrix-Valued Query [article]

Thee Chanyaswad, Alex Dytso, H. Vincent Poor, Prateek Mittal
2018 arXiv   pre-print
Particularly, we propose a novel differential privacy mechanism called the Matrix-Variate Gaussian (MVG) mechanism, which adds a matrix-valued noise drawn from a matrix-variate Gaussian distribution.  ...  In this work, we consider the design of differential privacy mechanism specifically for a matrix-valued query function.  ...  Clearly, the direction-derivation process via SVD needs to be private. Fortunately, there have been many works on differentially-private SVD [81, 82, 26] .  ... 
arXiv:1802.10077v1 fatcat:n7hvc4myzffzlgtyvqc7brwa6y

Efficient Public Good Provision in Networks: Revisiting the Lindahl Solution

Anil Jain
2017 International Finance Discussion Paper  
This completes the proof the centrality-stable a ∈ R n >0 is unique. EFFICIENT PUBLIC GOOD PROVISION IN NETWORKS: REVISITING THE LINDAHL SOLUTION 26 A.2. Proof of Proposition 2.  ...  The game is one of complete and symmetric information.  ...  Writing out the equilibrium condition in the new notation gives us: Differentiating this function with respect to θ gives: EFFICIENT PUBLIC GOOD PROVISION IN NETWORKS: REVISITING THE LINDAHL SOLUTION  ... 
doi:10.17016/ifdp.2017.1210 fatcat:prfzsm223vhsndnbs2qynwg3de

A Private and Computationally-Efficient Estimator for Unbounded Gaussians [article]

Gautam Kamath, Argyris Mouzakis, Vikrant Singhal, Thomas Steinke, Jonathan Ullman
2021 arXiv   pre-print
The primary new technical tool in our algorithm is a new differentially private preconditioner that takes samples from an arbitrary Gaussian 𝒩(0,Σ) and returns a matrix A such that A Σ A^T has constant  ...  We give the first polynomial-time, polynomial-sample, differentially private estimator for the mean and covariance of an arbitrary Gaussian distribution 𝒩(μ,Σ) in ℝ^d.  ...  Algorithm 3: Differentially Private CoarsePreconditioner , , , , ˆ ( ) Input: Samples 1 , . . . , ∈ R . Parameters , , , > 0, ˆ ≥ 0. Output: Matrix ∈ R × . Set 1 − ← ˆ .  ... 
arXiv:2111.04609v1 fatcat:6hglzegrxnaq3afqsikyf5hyoy

Private Alternating Least Squares: Practical Private Matrix Completion with Tighter Rates [article]

Steve Chien, Prateek Jain, Walid Krichene, Steffen Rendle, Shuang Song, Abhradeep Thakurta, Li Zhang
2021 arXiv   pre-print
We study the problem of differentially private (DP) matrix completion under user-level privacy.  ...  We design a joint differentially private variant of the popular Alternating-Least-Squares (ALS) method that achieves: i) (nearly) optimal sample complexity for matrix completion (in terms of number of  ...  Differentially private matrix completion revisited. In International Con- ference on Machine Learning, pp. 2215-2224. PMLR, 2018. Kearns, M., Pai, M., Roth, A., and Ullman, J.  ... 
arXiv:2107.09802v1 fatcat:ao7wdfub45bpxow736paoy72wm

Fast Dimension Independent Private AdaGrad on Publicly Estimated Subspaces [article]

Peter Kairouz, Mónica Ribero, Keith Rush, Abhradeep Thakurta
2021 arXiv   pre-print
We revisit the problem of empirical risk minimziation (ERM) with differential privacy.  ...  Our utility guarantee for the private ERM problem follows as a corollary to the regret guarantee of noisy AdaGrad.  ...  Introduction In this paper we revisit the problem of private empirical risk minimziation (ERM) with differential privacy [CMS11, BST14, SCS13, ACG + 16, BFTT19, MRTZ17, WLK + 17, INS + 19, PSY + 19, TAM19  ... 
arXiv:2008.06570v2 fatcat:sxe4w6juuzfo3m6ha2mvwquzca

Subspace Differential Privacy [article]

Jie Gao, Ruobin Gong, Fang-Yi Yu
2021 arXiv   pre-print
We discuss two design frameworks that convert well-known differentially private mechanisms, such as the Gaussian and the Laplace mechanisms, to subspace differentially private ones that respect the invariants  ...  Subspace differentially private mechanisms rid the need for post-processing due to invariants, preserve transparency and statistical intelligibility of the output, and can be suitable for distributed implementation  ...  Gaussian is a location-scale family completely characterized by its mean vector and covariance matrix.  ... 
arXiv:2108.11527v1 fatcat:wzmg5bfcqndffl5l4uashpkkye

Oneshot Differentially Private Top-k Selection [article]

Gang Qiao, Weijie J. Su, Li Zhang
2021 arXiv   pre-print
We show that the oneshot Laplace mechanism with a noise level of O(√(k)/) is approximately differentially private.  ...  Being able to efficiently and accurately select the top-k elements with differential privacy is an integral component of various private data analysis tasks.  ...  Preliminaries Before continuing, we pause to revisit some basic concepts in differential privacy. Definition 1.1.  ... 
arXiv:2105.08233v2 fatcat:oaqkkbiibrepldwnnnlqzttamm
« Previous Showing results 1 — 15 out of 13,664 results