Filters








77 Hits in 5.0 sec

The Dirichlet Mechanism for Differential Privacy on the Unit Simplex [article]

Parham Gohari, Bo Wu, Matthew Hale, Ufuk Topcu
2019 arXiv   pre-print
In the former case, we derive expressions for the differential privacy level of privatizing a single vector within the unit simplex.  ...  We find the mechanism well-suited for inputs within the unit simplex because it always returns a privatized output that is also in the unit simplex.  ...  Then, we show that the Dirichlet mechanism satisfies probabilistic differential privacy for identity queries. By an identity query, we mean privatizing a single vector within the unit simplex.  ... 
arXiv:1910.00043v1 fatcat:skw7fjyswrgwzmbwp3yl7osb5e

Privacy-Preserving Kickstarting Deep Reinforcement Learning with Privacy-Aware Learners [article]

Parham Gohari, Bo Chen, Bo Wu, Matthew Hale, Ufuk Topcu
2021 arXiv   pre-print
Therefore, we use the framework of differential privacy to develop a mechanism that securely shares the teacher's demonstrations with the student.  ...  The mechanism allows for the teacher to decide upon the accuracy of its demonstrations with respect to the privacy budget that it consumes, thereby granting the teacher full control over its data privacy  ...  The Dirichlet mechanism may achieve a higher accuracy than the conventional differential-privacy mechanisms because it does not require additional projections back onto the unit simplex.  ... 
arXiv:2102.09599v2 fatcat:5d6yfwhfifgjhjn7mjknvzm6ky

Privacy-Aware Load Ensemble Control: A Linearly-Solvable MDP Approach [article]

Ali Hassan, Deepjyoti Deka, Yury Dvorkin
2021 arXiv   pre-print
To this end, the concept of differential privacy is internalized into the MDP routine to protect transition probabilities and, thus, privacy of DR participants.  ...  The proposed approach also provides a trade-off between solution optimality and privacy guarantees, and is analyzed using real-world data from DR events in the New York University microgrid in New York  ...  This is achieved by leveraging the Dirichlet mechanism, which preserves unit simplex (i.e., the sum of transition probabilities for a given state is equal to one), and its coupling with the two variances  ... 
arXiv:2103.10828v2 fatcat:cbjwlttmefaqpmnjfzcre6q6zm

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.  ...  Finally, numerical experiments on two example environments validate the established relationship between the cost of privacy and the strength of data privacy protections.  ...  of the unit simplex to differ in only one entry because their components must sum to one.  ... 
arXiv:2004.07778v1 fatcat:b5q6x2wtgje7pmlkeyxc6qsoba

Privacy-Preserving Fraud Detection via Cooperative Mobile Carriers with Improved Accuracy

Wenyan Yao, Na Ruan, Feifan Yu, Weijia Jia, Haojin Zhu
2017 2017 14th Annual IEEE International Conference on Sensing, Communication, and Networking (SECON)  
To deal with this weakness, we also demonstrate that our method can detect the fraudulent accounts without leaking the private records and data of user accounts based on the differential privacy.  ...  We introduce the Latent Dirichlet Allocation (LDA) model to profile users in different carriers.  ...  For a query : ≤ , the mechanism that adds independently generated noise with distribution (0, ) : ( = ) = 1 2 (− || || 2 ) (20) it gives △ -differential privacy. Its important to note that ∑ , .  ... 
doi:10.1109/sahcn.2017.7964943 dblp:conf/secon/YaoRYJZ17 fatcat:p3jlvrommrhtvijrkehfrwmjc4

A Latent Class Modeling Approach for Generating Synthetic Data and Making Posterior Inferences from Differentially Private Counts [article]

Michelle Pistner Nixon, Andrés F. Barrientos, Jerome P. Reiter, Aleksandra Slavković
2022 arXiv   pre-print
We present a latent class modeling approach for post-processing differentially private marginal counts that can be used (i) to create differentially private synthetic data from the set of marginal counts  ...  Several algorithms exist for creating differentially private counts from contingency tables, such as two-way or three-way marginal counts.  ...  Acknowledgment This work was supported by the Pennsylvania State University, Duke University, ACI under award 1443014, and the National Science Foundation under an IGERT award #DGE-1144860, Big Data Social  ... 
arXiv:2201.10545v1 fatcat:fbruzaf6p5c2jlbebsodasilhi

Balancing Privacy-Utility of Differential Privacy Mechanism: A Collaborative Perspective

Hai Liu, Changgen Peng, Youliang Tian, Shigong Long, Zhenqiang Wu, Shahram Babaie
2021 Security and Communication Networks  
Differential privacy mechanism can maintain privacy-utility monotonicity. Thus, differential privacy mechanism does not obtain privacy-utility balance for numerical data.  ...  First, we constructed the collaborative model achieving privacy-utility balance of differential privacy mechanism.  ...  [16] introduced the Dirichlet mechanism with differential privacy, which is used for privatizing data inputs that belong to the unit simplex. e Dirichlet mechanism establishes a tradeoff between the  ... 
doi:10.1155/2021/5592191 fatcat:pep7qxl3pjgrvg5eauymlriz6m

Modeling and detecting anomalous topic access

Siddharth Gupta, Casey Hanson, Carl A Gunter, Mario Frank, David Liebovitz, Bradley Malin
2013 2013 IEEE International Conference on Intelligence and Security Informatics  
Specifically, we utilize a combination of Latent Dirichlet Allocation (LDA), for feature extraction, a k-nearest neighbor (k-NN) algorithm for outlier detection and evaluate the ability to identify different  ...  We argue that this model is appropriate for a meaningful range of attacks and develop a system based on topic summarization that is able to formalize the model and provide anomalous user detection effectively  ...  The authors thank You Chen of Vanderbilt University for insightful discussions during this research.  ... 
doi:10.1109/isi.2013.6578795 dblp:conf/isi/GuptaHGFLM13 fatcat:34x4penjvffzfju2qihdihkgb4

On the Theory and Practice of Privacy-Preserving Bayesian Data Analysis [article]

James Foulds, Joseph Geumlek, Max Welling, Kamalika Chaudhuri
2016 arXiv   pre-print
We show that a simple alternative based on the Laplace mechanism, the workhorse of differential privacy, is as asymptotically efficient as non-private posterior inference, under general assumptions.  ...  Bayesian inference has great promise for the privacy-preserving analysis of sensitive data, as posterior sampling automatically preserves differential privacy, an algorithmic notion of data privacy, under  ...  We also thank Mijung Park, Eric Nalisnick, and Babak Shahbaba for helpful discussions.  ... 
arXiv:1603.07294v2 fatcat:yrdgavap4radbjle5wojrevlra

Reduction and decomposition of differential automata: Theory and applications [chapter]

Alexey S. Matveev, Andrey V. Savkin
1998 Lecture Notes in Computer Science  
The ideal class group, finiteness of the ideal class group, Dirichlet units theorem.Texts / References: K. Ireland and M.  ...  Simulation of Mechanical Systems. SIMULINK (and/or similar package) based experiments on graphical synthesis of systems; Control synthesis for typical dynamical systems.  ... 
doi:10.1007/3-540-64358-3_48 fatcat:hqwvar3zbfftdhg4cybpjwogde

Differentially Private Learning of Undirected Graphical Models using Collective Graphical Models [article]

Garrett Bernstein, Ryan McKenna, Tao Sun, Daniel Sheldon, Michael Hay, Gerome Miklau
2017 arXiv   pre-print
We show that the approach of releasing noisy sufficient statistics using the Laplace mechanism achieves a good trade-off between privacy, utility, and practicality.  ...  We investigate the problem of learning discrete, undirected graphical models in a differentially private way.  ...  Acknowledgments This material is based upon work supported by the National Science Foundation under Grant Nos. 1409125, 1409143, 1421325, and 1617533.  ... 
arXiv:1706.04646v1 fatcat:jlh2qd2wmbgszgtknqsx7kkgkq

Learner-Private Convex Optimization [article]

Jiaming Xu, Kuang Xu, Dana Yang
2021 arXiv   pre-print
Our proofs learn on tools from the theory of Dirichlet processes, as well as a novel strategy designed for measuring information leakage under a full-gradient oracle.  ...  The repeated interactions, however, expose the learner to privacy risks from eavesdropping adversaries that observe the submitted queries.  ...  Acknowledgment The authors thank Niva Ran and Benjamin Ran for suggesting ideas that inspired the algorithm used in the upper bound of the Bayesian formulation of the problem.  ... 
arXiv:2102.11976v2 fatcat:3zgdmefk6jdq5ol55v5ksahtx4

Providing Access to Confidential Research Data Through Synthesis and Verification: An Application to Data on Employees of the U.S. Federal Government [article]

Andrés F. Barrientos, Alexander Bolton, Tom Balmat, Jerome P. Reiter, John M. de Figueiredo, Ashwin Machanavajjhala, Yan Chen, Charley Kneifel, Mark DeLong
2018 arXiv   pre-print
We also present novel verification algorithms for regression coefficients that satisfy differential privacy.  ...  The analysis on the confidential data reveals pay differentials across races not documented in published studies.  ...  For Z|G, we create a one-to-one function T G to map Z into a space of permutations dependent on G. We model T G (Z)|G using a latent model defined on the simplex space.  ... 
arXiv:1705.07872v2 fatcat:bqenh3aopve7focahkan2edjku

Statistical Inference in the Differential Privacy Model [article]

Huanyu Zhang
2021 arXiv   pre-print
This thesis is a summary of the author's several works during his Ph.D. Besides, it has established the optimal sample complexity of differentially private closeness testing.  ...  Theorem 2 is a bound on the risk for pure differential privacy (δ = 0).  ...  The case when δ = 0 is called pure differential privacy. For simplicity, we denote pure differential privacy as ε-differential privacy (ε-DP). Next we show two properties of differential privacy.  ... 
arXiv:2108.05000v1 fatcat:5o2dopkueja7hgi5p7ookyy3ru

A Brief History of Statistical Models for Network Analysis and Open Challenges

Stephen E. Fienberg
2012 Journal of Computational And Graphical Statistics  
and by grant FA9550-12-1-0392 from the U.S.  ...  ACKNOWLEDGMENTS This research was supported by the Singapore National Research Foundation under its International Research Centre @ Singapore Funding Initiative and administered by the IDM Programme Office  ...  MCMC is the mechanism for estimating the posterior distribution.  ... 
doi:10.1080/10618600.2012.738106 fatcat:gfnv7itsmjg6jfhrbggmyudow4
« Previous Showing results 1 — 15 out of 77 results