Filters








15,561 Hits in 2.4 sec

Differential Privacy on Finite Computers [article]

Victor Balcer, Salil Vadhan
2019 arXiv   pre-print
We consider the problem of designing and analyzing differentially private algorithms that can be implemented on discrete models of computation in strict polynomial time, motivated by known attacks on floating  ...  privacy guarantee.  ...  Acknowledgments We thank the Harvard Privacy Tools differential privacy research group, particularly Mark Bun and Kobbi Nissim, for informative discussions and feedback.  ... 
arXiv:1709.05396v3 fatcat:7byfnhk4fnerxbtk72ismmxnjy

Differential Privacy on Finite Computers

Victor Balcer, Salil Vadhan
2019 Journal of Privacy and Confidentiality  
We consider the problem of designing and analyzing differentially private algorithms that can be implemented on discrete models of computation in strict polynomial time, motivated by known attacks on floating  ...  privacy guarantee.  ...  We thank the Harvard Privacy Tools differential privacy research group, particularly Mark Bun and Kobbi Nissim, for informative discussions and feedback.  ... 
doi:10.29012/jpc.679 fatcat:5xdqegfd3zdiriks4tavhgyrbu

Automated Methods for Checking Differential Privacy [article]

Gilles Barthe, Rohit Chadha, Vishal Jagannath, A. Prasad Sistla and Mahesh Viswanathan
2019 arXiv   pre-print
Differential privacy is a de facto standard for statistical computations over databases that contain private data.  ...  We propose the first decision procedure for checking the differential privacy of a non-trivial class of probabilistic computations.  ...  Introduction Differential privacy [16] is a gold standard for privacy of statistical computations.  ... 
arXiv:1910.04137v1 fatcat:h2qddvpih5eylez7c2qv7zj2bq

Differential Privacy as a Mutual Information Constraint

Paul Cuff, Lanqing Yu
2016 Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security - CCS'16  
Intuitively, one must come to terms with what differential privacy does and does not guarantee.  ...  The mutual-information differential privacy is in fact sandwiched between ϵ-differential privacy and (ϵ,δ)-differential privacy in terms of its strength.  ...  If k queries Y q1 , Y q2 , …, Y qk each have differential privacy ε and are conditionally independent, the combined they have kε privacy.  ... 
doi:10.1145/2976749.2978308 dblp:conf/ccs/CuffY16 fatcat:teiovmawzjcxxjrtajaezkqyda

Asymmetric Distances for Approximate Differential Privacy

Dmitry Chistikov, Andrzej S. Murawski, David Purser, Michael Wagner
2019 International Conference on Concurrency Theory  
Differential privacy is a widely studied notion of privacy for various models of computation, based on measuring differences between probability distributions.  ...  We confirm this using several case studies, where we also demonstrate, on a randomised response mechanism, that the estimates based on ld α can beat standard differential privacy composition theorems.  ...  When one cannot expect to achieve this pure -differential privacy, the relaxed approximate differential privacy is used [24] .  ... 
doi:10.4230/lipics.concur.2019.10 dblp:conf/concur/ChistikovMP19 fatcat:4o7ysinmgfgs5clwsayinqh7ry

Improving Deep Learning with Differential Privacy using Gradient Encoding and Denoising [article]

Milad Nasr, Reza Shokri, Amir houmansadr
2020 arXiv   pre-print
Previous work has investigated training models with differential privacy (DP) guarantees through adding DP noise to the gradients.  ...  We show that our mechanism outperforms the state-of-the-art DPSGD; for instance, for the same model accuracy of 96.1% on MNIST, our technique results in a privacy bound of ϵ=3.2 compared to ϵ=6 of DPSGD  ...  Milad Nasr is supported by a Google PhD Fellowship in Security and Privacy.  ... 
arXiv:2007.11524v1 fatcat:znuru2fcdngw3nvggazy6lwpbq

Editorial for Volume 9 Issue 2

Jonathan Ullman, Lars Vilhuber
2019 Journal of Privacy and Confidentiality  
from the 3rd Workshop on Theory and Practice of Differential Privacy, which was held in Dallas, TX on October 30, 2017 as part of the ACM Conference on Computer Security (CCS).  ...  Researchers in differential privacy span many distinct research communities, including algorithms, computer security, cryptography, databases, data mining, machine learning, statistics, programming languages  ...  30, 2017 as part of the ACM Conference on Computer Security (CCS).  ... 
doi:10.29012/jpc.731 fatcat:ov4voneko5fabmc23kelmllyzq

Towards Differential Privacy for Symbolic Systems [article]

Austin Jones, Kevin Leahy, Matthew Hale
2018 arXiv   pre-print
This work uses the framework of differential privacy, which is a statistical notion of privacy that makes it unlikely that privatized data will reveal anything meaningful about underlying sensitive data  ...  To bring differential privacy to symbolic control systems, we develop an exponential mechanism that approximates a sensitive word using a randomly chosen word that is likely to be near it.  ...  Section II presents the relevant background on Levenshtein automata and differential privacy. Then, Section III formally states the differential privacy problem that is the focus of the paper.  ... 
arXiv:1809.08634v1 fatcat:2cl4n64tq5cy5pwhwqvmyftwaa

Preserving differential privacy under finite-precision semantics

Ivan Gazeau, Dale Miller, Catuscia Palamidessi
2013 Electronic Proceedings in Theoretical Computer Science  
We focus here on differential privacy, an approach to privacy that emerged from the area of statistical databases and is now widely applied also in other domains.  ...  Finally, we illustrate our results on two cases of noise-generating distributions: the standard Laplacian mechanism commonly used in differential privacy, and a bivariate version of the Laplacian recently  ...  Our main concern is to establish a bound on the degradation of privacy induced by both the finite representation and by the computational errors in the generation of the noise.  ... 
doi:10.4204/eptcs.117.1 fatcat:26i3nzbkmfcfbblkoqk56wjseu

Preserving differential privacy under finite-precision semantics

Ivan Gazeau, Dale Miller, Catuscia Palamidessi
2016 Theoretical Computer Science  
We focus here on differential privacy, an approach to privacy that emerged from the area of statistical databases and is now widely applied also in other domains.  ...  The approximation introduced by finite-precision representation of continuous data can induce arbitrarily large information leaks even when the computation using exact semantics is secure.  ...  Our main concern is to establish a bound on the degradation of privacy induced by both the finite representation and by the computational errors in the generation of the noise.  ... 
doi:10.1016/j.tcs.2016.01.015 fatcat:2quznsndkveszlqolzc26nyhpu

Privacy-Preserving Policy Synthesis in Markov Decision Processes [article]

Parham Gohari, Matthew Hale, Ufuk Topcu
2020 arXiv   pre-print
We use differential privacy as the mathematical definition of privacy. The algorithm first perturbs the transition probabilities using a mechanism that provides differential privacy.  ...  We also show that computing the cost of privacy has time complexity that is polynomial in the parameters of the problem.  ...  For finite-horizon MDPs, we show that we can compute the cost of privacy in polynomial time via a backward-in-time recursive algorithm.  ... 
arXiv:2004.07778v1 fatcat:b5q6x2wtgje7pmlkeyxc6qsoba

Finite Sample Differentially Private Confidence Intervals

Vishesh Karwa, Salil Vadhan, Marc Herbstritt
2018 Innovations in Theoretical Computer Science  
We study the problem of estimating finite sample confidence intervals of the mean of a normal population under the constraint of differential privacy.  ...  We also prove lower bounds on the expected size of any differentially private confidence set showing that our the parameters are optimal up to polylogarithmic factors.  ...  We are grateful to the Harvard Privacy Tools differential privacy research group for illuminating and motivating discussions, particularly James Honaker, Gary King, Kobbi Nissim and Uri Stemmer.  ... 
doi:10.4230/lipics.itcs.2018.44 dblp:conf/innovations/KarwaV18 fatcat:bcea5y5kqrhgvfe5fmf622z5mu

Differential Private Trajectory Obfuscation [chapter]

Roland Assam, Marwan Hassani, Thomas Seidl
2013 Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering  
We propose a novel technique to ensure location privacy for mobility data using differential privacy.  ...  We conducted numerous experiments on real and synthetic datasets, and show that our approach produces superior privacy results when compared to state-of-the-art techniques.  ...  This research was funded by the cluster of excellence on Ultra-high speed Mobile Information and Communication (UMIC) of the DFG (German Research Foundation grant EXC 89).  ... 
doi:10.1007/978-3-642-40238-8_12 fatcat:no6uy7fskjgwjjry4jl3w35jlm

Differentially Private Reinforcement Learning with Linear Function Approximation

Xingyu Zhou
2022 Abstract Proceedings of the 2022 ACM SIGMETRICS/IFIP PERFORMANCE Joint International Conference on Measurement and Modeling of Computer Systems  
Compared to existing private RL algorithms that work only on tabular finite-state, finiteactions MDPs, we take the first step towards privacy-preserving learning in MDPs with large state and action spaces  ...  Markov decision processes (MDPs) under the constraints of differential privacy (DP).  ...  For the general RL problem, there are only a few works that consider differential privacy [7, 19] .  ... 
doi:10.1145/3489048.3522648 fatcat:pq4hpup2ovc4zhkwji3ve44qr4

A framework for privacy preserving statistical analysis on distributed databases

Bing-Rong Lin, Ye Wang, Shantanu Rane
2012 2012 IEEE International Workshop on Information Forensics and Security (WIFS)  
Privacy of the marginal and joint statistics on the distributed databases is formulated using a new model called δ−distributional −differential privacy.  ...  Further, an adversary who looks at the published data must not even be able to compute statistical measures on it.  ...  Informally, differential privacy means that the result of a function computed on a database of respondents is almost insensitive to the presence or absence of a particular respondent.  ... 
doi:10.1109/wifs.2012.6412626 dblp:conf/wifs/LinWR12 fatcat:giu5z2vhvrgkphowyxk6ncm7fi
« Previous Showing results 1 — 15 out of 15,561 results