Filters








13,318 Hits in 15.5 sec

A Statistical Threshold for Adversarial Classification in Laplace Mechanisms [article]

Ayşe Ünsal, Melek Önen
2021 arXiv   pre-print
Finally, we adapt the Kullback-Leibler differential privacy to adversarial classification.  ...  Corresponding error probabilities are analytically derived and ROC curves are presented for various levels of the sensitivity, the absolute mean of the attack and the privacy parameter.  ...  The authors wish to continue this line of work on different differentially private mechanisms (e.g.  ... 
arXiv:2105.05610v4 fatcat:2qxohjcst5ghvdvomwwesbc6vi

A Statistical Threshold for Adversarial Classification in Laplace Mechanisms

Ayse Unsal, Melek Onen
2021 2021 IEEE Information Theory Workshop (ITW)  
Thus, there are two sides of what the adversary wants to achieve: (i) s/he gives false data by modifying the released information with the biggest possible difference from the real data, (ii) all this  ...  Thus, we establish statistical thresholds for detecting the attack as a function of the error probabilities for Laplace mechanisms through one-sided and twosided hypothesis tests.  ... 
doi:10.1109/itw48936.2021.9611472 fatcat:kamawdvnzjawbpesthn2rtx6ua

Optimal Obfuscation Mechanisms via Machine Learning [article]

Marco Romanelli and Konstantinos Chatzikokolakis and Catuscia Palamidessi
2020 arXiv   pre-print
We apply our method to the case of location privacy, and we perform experiments on synthetic data and on real data from the Gowalla dataset.  ...  We evaluate the privacy of the mechanism not only by its capacity to defeat the classifier, but also in terms of the Bayes error, which represents the strongest possible adversary.  ...  (ε, δ)-differential privacy.  ... 
arXiv:1904.01059v5 fatcat:s7fz5767fbailpfkdtfcnmfo3i

Protecting Global Properties of Datasets with Distribution Privacy Mechanisms [article]

Michelle Chen, Olga Ohrimenko
2022 arXiv   pre-print
Alongside the rapid development of data collection and analysis techniques in recent years, there is increasingly an emphasis on the need to address information leakage associated with such usage of data  ...  privacy baseline.  ...  Nonetheless, Figure 3 shows the error incurred by Gaussian variants of the Expected Value Mechanisms as Δ𝑝 varies, with the error incurred by differential privacy mechanisms shown for comparison.  ... 
arXiv:2207.08367v1 fatcat:pmjdwgt77fb5nm4ynclflcoryu

Not All Attributes are Created Equal: d_X-Private Mechanisms for Linear Queries [article]

Parameswaran Kamalaruban and Victor Perrier and Hassan Jameel Asghar and Mohamed Ali Kaafar
2019 arXiv   pre-print
Differential privacy provides strong privacy guarantees simultaneously enabling useful insights from sensitive datasets.  ...  We describe a systematic procedure to tailor any existing differentially private mechanism that assumes a query set and a sensitivity vector as input into its d_X-private variant, specifically focusing  ...  Privacy is provided for the n = 10, 000 subjects who can be in any of the N locations in the data universe, with higher privacy for nearby locations.  ... 
arXiv:1806.02389v2 fatcat:ugp56avfgrcppncueozehiqgqe

Privacy and Mechanism Design [article]

Mallesh Pai, Aaron Roth
2013 arXiv   pre-print
Here, we survey several facets of this study, and differential privacy plays a role in more than one way.  ...  This paper is a survey of recent work at the intersection of mechanism design and privacy.  ...  Can differential privacy and its variants assist in the design of such information release?  ... 
arXiv:1306.2083v1 fatcat:rpgbfnwaljeknjhjivwcrldnzm

Privacy and mechanism design

Mallesh M. Pai, Aaron Roth
2013 ACM SIGecom Exchanges  
Here, we survey several facets of this study, and differential privacy plays a role in more than one way.  ...  This paper is a survey of recent work at the intersection of mechanism design and privacy.  ...  Can differential privacy and its variants assist in the design of such information release?  ... 
doi:10.1145/2509013.2509016 fatcat:m5hvdv2pg5gfvbfs2gknjzpasq

KNG: The K-Norm Gradient Mechanism [article]

Matthew Reimherr, Jordan Awan
2021 arXiv   pre-print
This paper presents a new mechanism for producing sanitized statistical summaries that achieve differential privacy, called the K-Norm Gradient Mechanism, or KNG.  ...  In addition to theoretical guarantees on privacy and utility, we confirm the utility of KNG empirically in the settings of linear and quantile regression through simulations.  ...  Because the KNG results in a location family in this case, the integrating constant does not depend on the data.  ... 
arXiv:1905.09436v2 fatcat:vpjahcx45bbqdi7doh3m2rcvtu

Not All Attributes are Created Equal: dX -Private Mechanisms for Linear Queries

Parameswaran Kamalaruban, Victor Perrier, Hassan Jameel Asghar, Mohamed Ali Kaafar
2020 Proceedings on Privacy Enhancing Technologies  
Differential privacy provides strong privacy guarantees simultaneously enabling useful insights from sensitive datasets.  ...  We describe a systematic procedure to tailor any existing differentially private mechanism that assumes a query set and a sensitivity vector as input into its dX -private variant, specifically focusing  ...  Privacy is provided for the n = 10, 000 subjects who can be in any of the N locations in the data universe, with higher privacy for nearby locations.  ... 
doi:10.2478/popets-2020-0007 dblp:journals/popets/KamalarubanPAK20 fatcat:vmdcveafofb4fcgf5zpybs7ulq

Rényi Differential Privacy Mechanisms for Posterior Sampling [article]

Joseph Geumlek, Shuang Song, Kamalika Chaudhuri
2017 arXiv   pre-print
Using a recently proposed privacy definition of R\'enyi Differential Privacy (RDP), we re-examine the inherent privacy of releasing a single sample from a posterior distribution.  ...  We exploit the impact of the prior distribution in mitigating the influence of individual data points.  ...  Differential Privacy (DP) [6] .  ... 
arXiv:1710.00892v1 fatcat:ragje7ffbvganm5255e5szyu3m

Toward a Comparison of Classical and New Privacy Mechanism

Daniel Heredia-Ductram, Miguel Nunez-del-Prado, Hugo Alatrista-Salas
2021 Entropy  
To fill this gap, we compare classical approaches of privacy techniques like Statistical Disclosure Control and Differential Privacy techniques to more recent techniques such as Generative Adversarial  ...  This phenomenon has raised privacy concerns around the globe, compelling states to enforce data protection laws.  ...  Data Sanitization through Differential Privacy Filters In this section, two techniques based on Differential Privacy mechanisms were applied to data at our disposal.  ... 
doi:10.3390/e23040467 pmid:33921188 fatcat:7p4rvbjuvfeypei3nplzt3d5bi

Generalized Gaussian Mechanism for Differential Privacy

Fang Liu
2018 IEEE Transactions on Knowledge and Data Engineering  
Practical applications of DP involve development of DP mechanisms to release results at a pre-specified privacy budget.  ...  affects the prediction power of a classifier constructed with the sanitized data in the adult experiment.  ...  Differential privacy (DP) provides a strong privacy guarantee to data release without making assumptions about the background knowledge or behavior of data users [1, 2, 3] .  ... 
doi:10.1109/tkde.2018.2845388 fatcat:qguntj7adbcgpagtgzgyu3t7oi

Differentially Private Binary- and Matrix-Valued Data Query: An XOR Mechanism

Tianxi Ji, Pan Li, Emre Yilmaz, Erman Ayday, Yanfang Ye, Jinyuan Sun
2021 Proceedings of the VLDB Endowment  
Differential privacy has been widely adopted to release continuousand scalar-valued information on a database without compromising the privacy of individual data records in it.  ...  Then, to generate the parameters in the matrix-valued Bernoulli distribution, we develop a heuristic approach to minimize the expected square query error rate under 𝜖-differential privacy constraint.  ...  for the proposed XOR mechanism to achieve 𝜖-differential privacy. • We analyze the utility of the XOR mechanism through the lens of expected square query error rate of a given query. • We devise a heuristic  ... 
dblp:journals/pvldb/Ji0YAYS21 fatcat:rgzgdobq4nc4djnun56kkwczzq

Benefits and Pitfalls of the Exponential Mechanism with Applications to Hilbert Spaces and Functional PCA [article]

Jordan Awan, Ana Kenney, Matthew Reimherr, Aleksandra Slavković
2019 arXiv   pre-print
The exponential mechanism is a fundamental tool of Differential Privacy (DP) due to its strong privacy guarantees and flexibility.  ...  We study its extension to settings with summaries based on infinite dimensional outputs such as with functional data analysis, shape analysis, and nonparametric statistics.  ...  Acknowledgments This research was supported in part by the following grants to Pennsylvania State University: NSF Grant SES-1534433, NSF Grant DMS-1712826, and NIH Grant 5T32LM012415-03 via the Biomedical Big Data  ... 
arXiv:1901.10864v1 fatcat:y6ef7xw35jfz3kqha6qqnvba5i

Structure and Sensitivity in Differential Privacy: Comparing K-Norm Mechanisms [article]

Jordan Awan, Aleksandra Slavkovic
2019 arXiv   pre-print
Differential privacy (DP), provides a framework for provable privacy protection against arbitrary adversaries, while allowing the release of summary statistics and synthetic data.  ...  We address the problem of releasing a noisy real-valued statistic vector T, a function of sensitive data under DP, via the class of K-norm mechanisms with the goal of minimizing the noise added to achieve  ...  The ∞ norm variant of the K-mech is proposed in , as a mechanism to optimize the worst case error when releasing one-way marginals of a high-dimensional binary database.  ... 
arXiv:1801.09236v3 fatcat:ffab5ea5kreufgji6jcslh4aiq
« Previous Showing results 1 — 15 out of 13,318 results