29,129 Hits in 2.6 sec

Flexible Accuracy for Differential Privacy [article]

Aman Bansal, Rahul Chunduru, Deepesh Data, Manoj Prabhakaran
2021 arXiv   pre-print
Our main contribution is in augmenting differential privacy with Flexible Accuracy, which allows small distortions in the input (e.g., dropping outliers) before measuring accuracy of the output, allowing  ...  Differential Privacy (DP) has become a gold standard in privacy-preserving data analysis.  ...  We present "composition theorems" which yield flexible accuracy and differential privacy guarantees for M in terms of those for M 1 and M 2 .  ... 
arXiv:2110.09580v1 fatcat:6hrekblz5bg2bfutupbwp5gcmi

Differential Privacy Based Access Control [chapter]

Nadia Metoui, Michele Bezzi
2016 Lecture Notes in Computer Science  
We propose a novel privacy-aware access control model, based on differential privacy.  ...  The model allows for data access at different privacy levels, generating an anonymized data set according to the privacy clearance of each request.  ...  Still, the accuracy level of 75% could be enough for many benchmarking tasks. The accuracy goes up, as expected, for higher privacy clearance values.  ... 
doi:10.1007/978-3-319-48472-3_61 fatcat:y2qvvp4uwfgzlcsxpa4fikhxpi

Editorial for Volume 9 Issue 2

Jonathan Ullman, Lars Vilhuber
2019 Journal of Privacy and Confidentiality  
also flexible enough to allow for a wide variety of data analyses to be performed with a high degree of utility.  ...  Differential privacy is a promising approach to privacy-preserving data analysis that provides strong worst-case guarantees about the harm that a user could suffer from contributing their data, but is  ...  a fixed privacy constraint at minimal cost in accuracy.  ... 
doi:10.29012/jpc.731 fatcat:ov4voneko5fabmc23kelmllyzq

A privacy preserving system for friend locator applications

Bin Zan, Tingting Sun, Marco Gruteser, Fei Hu, Yanyong Zhang
2011 Proceedings of the 9th ACM international symposium on Mobility management and wireless access - MobiWac '11  
Additionally, we use the polygon decomposition method to achieve both accuracy and flexibility especially for irregular areas of interest.  ...  Finally, through numerical analysis and simulation, we show that the proposed system and algorithm can achieve high privacy, efficiency, accuracy and flexibility.  ...  In this paper, we develop a new privacy preserving system and algorithm to achieve high privacy, accuracy, efficiency and flexibility for friend locator applications.  ... 
doi:10.1145/2069131.2069147 dblp:conf/mobiwac/ZanSGHZ11 fatcat:uhxsbptobfg7fhfv7ej2xu7ypi

Local Differential Privacy for Deep Learning

Pathum Chamikara Mahawaga Arachchige, Peter Bertok, Ibrahim Khalil, Dongxi Liu, Seyit Camtepe, Mohammed Atiquzzaman
2019 IEEE Internet of Things Journal  
Hence, the randomization module can operate as an NFV privacy preservation service in an SDN-controlled NFV, making LATENT more practical for IoT-driven cloud-based environments compared to existing approaches  ...  Our experimental evaluation of LATENT on convolutional deep neural networks demonstrates excellent accuracy (e.g. 91 model quality even under low privacy budgets (e.g. ε=0.5).  ...  We achieve 95%-96% testing accuracy and 90%-91% testing accuracy for the MNIST dataset and CIFAR-10 dataset, respectively, with a high level of privacy (0.5-differential privacy).  ... 
doi:10.1109/jiot.2019.2952146 fatcat:ss4vbsqkh5d27cb7p3kpzuog5y

More Flexible Differential Privacy: The Application of Piecewise Mixture Distributions in Query Release [article]

David B. Smith, Kanchana Thilakarathna, Mohamed Ali Kaafar
2017 arXiv   pre-print
Then the relevant distributions are utilised to theoretically prove the privacy and accuracy guarantees of the proposed mechanisms.  ...  While differential privacy has attracted significant interest from academia and industry by providing rigorous and reliable privacy guarantees, the reduced utility and inflexibility of current differentially  ...  Differential privacy is preserved and the dataset curator is provided with more flexibility in mechanism design.  ... 
arXiv:1707.01189v3 fatcat:sugcxewma5fifni27ekvflqapu

Privacy Enhancement via Dummy Points in the Shuffle Model [article]

Xiaochen Li, Weiran Liu, Hanwen Feng, Kunzhe Huang, Jinfei Liu, Kui Ren, Zhan Qin
2021 arXiv   pre-print
We propose DUMP (DUMmy-Point-based), a framework for privacy-preserving histogram estimation in the shuffle model.  ...  The shuffle model is recently proposed to address the issue of severe utility loss in Local Differential Privacy (LDP) due to distributed data randomization.In the shuffle model, a shuffler is utilized  ...  Particularly, for Ratings, [6] cannot even achieve ( , )-privacy with ≤ 1. -Accuracy.  ... 
arXiv:2009.13738v2 fatcat:tdysrn2nxjbjjkkbgcbyn2vbwm

FLDP: Flexible strategy for local differential privacy [article]

Dan Zhao, Suyun Zhao, Ruixuan Liu, Cuiping Li, Wenjuan Liang, Hong Chen
2022 arXiv   pre-print
Local differential privacy (LDP), a technique applying unbiased statistical estimations instead of real data, is often adopted in data collection.  ...  First, we present this weakened but flexible LDP (FLDP) notion. We then prove the association with LDP and DP. Second, we design an FHR approach for the common FO issue while satisfying FLDP.  ...  Differential privacy as the de facto standard for private data release was first introduced by Dwork [1] .  ... 
arXiv:2203.14875v1 fatcat:utxeungm6begfptj52wswlhbyy

Towards Differentially Private Text Representations [article]

Lingjuan Lyu, Yitong Li, Xuanli He, Tong Xiao
2020 arXiv   pre-print
For the randomization module, we propose a novel local differentially private (LDP) protocol to reduce the impact of privacy parameter ϵ on accuracy, and provide enhanced flexibility in choosing randomization  ...  probabilities for LDP.  ...  Without an untrusted server, Shokri and Shmatikov [17] proposed to blur local model gradients by adding noise using differential privacy.  ... 
arXiv:2006.14170v1 fatcat:pk6hsg2kvzdhrptufq5pp2tm5e

Using Feature Selection to Improve the Utility of Differentially Private Data Publishing

Yasser Jafer, Stan Matwin, Marina Sokolova
2014 Procedia Computer Science  
Within exiting privacy models, differential privacy is considered one of the strongest privacy protection techniques that does not make any assumption about the attacker's background knowledge.  ...  As a result, when noise is added to satisfy differential privacy, its distorting effect is minimized and high utility of the data is preserved.  ...  Acknowledgement The authors acknowledge the support for their work provided by the Natural Sciences and Engineering Research Council of Canada through the Strategic Grant Program, as well as the support  ... 
doi:10.1016/j.procs.2014.08.076 fatcat:wsut4jmtrjg2lhdoprwzuls524

dpart: Differentially Private Autoregressive Tabular, a General Framework for Synthetic Data Generation [article]

Sofiane Mahiou, Kai Xu, Georgi Ganev
2022 arXiv   pre-print
We propose a general, flexible, and scalable framework dpart, an open source Python library for differentially private synthetic data generation.  ...  PrivBayes seems to capture the marginals better than dp-synthpop for all privacy budgets. Accuracy.  ...  Accepted at the Theory and Practice of Differential Privacy (TPDP) 2022, part of ICML 2022.  ... 
arXiv:2207.05810v1 fatcat:rmck2gukbbd6bamjltd5vgcqhe

Practical Privacy Filters and Odometers with Rényi Differential Privacy and Applications to Differentially Private Deep Learning [article]

Mathias Lécuyer
2021 arXiv   pre-print
We demonstrate two applications of this theorem for DP deep learning: adapting the noise or batch size online to improve a model's accuracy within a fixed total privacy loss, and stopping early when fine-tuning  ...  However, existing composition theorems present a tension between efficiency and flexibility.  ...  Due to DP's strong privacy guarantees, multiple efforts exist to integrate it into libraries, including for statistical queries (Google Differential Privacy; OpenDP), machine learning (IBM Differential  ... 
arXiv:2103.01379v2 fatcat:xiq554pw35axph3g2xkh3mejhe

Enhancing the Privacy of Federated Learning with Sketching [article]

Zaoxing Liu, Tian Li, Virginia Smith, Vyas Sekar
2019 arXiv   pre-print
Our initial findings show that it is possible to provide strong privacy guarantees for federated learning without sacrificing performance or accuracy.  ...  Existing efforts that aim to improve the privacy of federated learning make compromises in one or more of the following key areas: performance (particularly communication cost), accuracy, or privacy.  ...  Flexible, private federated learning systems.  ... 
arXiv:1911.01812v1 fatcat:mquqgd2ykjepdidtf5bx4rkkpq

Differentially Private Generative Adversarial Networks for Time Series, Continuous, and Discrete Open Data [article]

Lorenzo Frigerio, Anderson Santana de Oliveira, Laurent Gomez, Patrick Duverger
2019 arXiv   pre-print
This paper aims at creating a framework for releasing new open data while protecting the individuality of the users through a strict definition of privacy called differential privacy.  ...  However, it is always difficult to create new high-quality datasets with the required privacy guarantees for many use cases.  ...  Classification accuracy for training sets generated by different models Fig. 4 . classification accuracy Average for 5 runs for different noise values .  ... 
arXiv:1901.02477v2 fatcat:t4xpvk7smjfc7ormvq6negozbm

Assessing differentially private deep learning with Membership Inference [article]

Daniel Bernau, Philip-William Grassal, Jonas Robl, Florian Kerschbaum
2020 arXiv   pre-print
This suggests that local differential privacy is a sound alternative to central differential privacy for differentially private deep learning, since small ϵ in central differential privacy and large ϵ  ...  We empirically compare local and central differential privacy mechanisms under white- and black-box membership inference to evaluate their relative privacy-accuracy trade-offs.  ...  Acknowledgements We thank Steffen Schneider for his instrumental contribution to implementation and analysis of white box MI attacks.  ... 
arXiv:1912.11328v4 fatcat:yscawmzefrhrbcf37rhavwq6vm
« Previous Showing results 1 — 15 out of 29,129 results