PrivBayes: private data release via bayesian networks

By Jun Zhang, Graham Cormode, Cecilia M. Procopiuc, Divesh Srivastava, Xiaokui Xiao. SIGMOD Conference, 2014

705 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Science Applications

View Paper: https://www.semanticscholar.org/paper/2faed3265a07adc15edf7eb157a34c08732b1101

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Science Applications

Citations in the corpus

Synthetic Data for Deep Learning (2019, methodology): “present the PrivBayes approach [691]: construct a Bayesian network that captures the correlations and dependencies between data attributes, inject noise into the marginals that constitute this network, and then sample from the perturbed network to produce the private synthetic dataset.”
PrivateSQL: A Differentially Private SQL Query Engine (2019, background): “This component is probably the most well understood as it is an instance of a common problem studied in the DP literature – answering a set of linear queries on a single table [22, 31, 39].”
Differentially Private Algorithms for Empirical Machine Learning (2014, background): “The sensitivity of information gain function has been shown to be O(log n) [9, 29], wheren is (an upper bound on) the number of tuples in the dataset.”
Collective Data-Sanitization for Preventing Sensitive Information Inference Attacks in Social Networks (2018, background): “both [17] and [18] sample a noise-data with differential privacy guarantees.”
Synthetic Data for Social Good (2017, methodology): “In that case the value is reset to 0 [24].”
Differentially Private High-Dimensional Data Publication via Sampling-Based Inference (2015, methodology): “We set θ = 4 (for θusefulness) to select the appropriate value of the degree of a Bayesian network, as recommended in [29].”
Understanding the Sparse Vector Technique for Differential Privacy (2016, neutral): “, [1, 13, 17, 18, 19], are in the non-interactive setting.”
Differential Privacy in Telco Big Data Platform (2015, background): “…on DP from industrial perspectives: • Implementation in telco big data platform: We broadly categorize the recent DP implementations for DTs into three basic architectures: 1) Data Publication Architecture [25, 39]; 2) Separated Architecture [7]; and 3) Hybridized Architecture [7, 13].”
How to DP-fy ML: A Practical Guide to Machine Learning with Differential Privacy (2023, methodology): “For example, Zhang et al. (2014) proposed a PrivBayes method that constructs a Bayesin network to model the interactions of the features.”
Differentially Private Mixture of Generative Neural Networks (2017, background): “2,14,26,34,35,51]. Alas, these can often support only the release of succinct data representations, such as histograms or contingency tables. Other mechanisms add noise directly to a generative model [9,32,33,59]. In this paper, we follow this approach, while, in a ﬁrst-of-its-kind attempt, focusing”
$\textsf{LoPub}$ : High-Dimensional Crowdsourced Data Publication With Local Differential Privacy (2016, methodology): “However, due to combinations, the possible domain will increase exponentially with the number of dimensions, thus leading to both low scalability and signal-noise-ratio problems [35].”
Modeling Tabular data using Conditional GAN (2019, methodology): “For simulated data from Gaussian mixture, CLBN and PrivBN suffer because continuous numeric data has to be discretized before modeling using Bayesian networks.”
EKTELO: A Framework for Defining Differentially-Private Computations (2018, methodology): “One of the few algorithms designed to scale to high dimensions is PrivBayes [47].”
Differential Privacy Techniques for Cyber Physical Systems: A Survey (2018, background): “Similarly, in the domain of big data, maintaining privacy protection along with issue of dimensionality because of large data volume and computation overhead is a big challenge for researchers [335], [336].”
Optimizing error of high-dimensional statistical queries under differential privacy (2018, methodology): “PrivBayes first privately fits a Bayes- ian network on the data and then generates a synthetic dataset by drawing samples from the Bayesian network.”
Differentially Private Data Publishing and Analysis: A Survey (2017, methodology): “[50] followed the above rationale and used a Bayesian network to deal with high dimensionality.”
Generating synthetic data in finance: opportunities, challenges and pitfalls (2020, background): “[22] present a synthetic data generator based on Bayesian networks that is also differentially private.”