The Optimal Noise-Adding Mechanism in Differential Privacy

By Quan Geng, Pramod Viswanath. 2012

194 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Science Applications

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

AdaCliP: Adaptive Clipping for Private SGD (2019, background): “Recently the optimal noise distributions to generate the least amount of noise have been proposed for both ( , 0) and ( , δ) differential privacy [18, 19].”
Protection Against Reconstruction and Its Applications in Private Federated Learning (2018, methodology): “1) provides a randomized-responsebased mechanism achieving order optimal scaling for the mean-squared error E[(Z − r)2], which is r2 maxe −2ε/3 for ε ≥ 1 (see Corollary 8 in [38]).”
Differentially Private Generative Adversarial Networks for Time Series, Continuous, and Discrete Open Data (2019, background): “Traditionally, many approaches tried to reach differential privacy by adding noise to the data in order to protect personal information [8,5,12], however they have never been able to provide satisfying results on real semantically-rich data; most of the implementations were limited to very specific ”
Optimal Noise-Adding Mechanism in Additive Differential Privacy (2018, methodology): “Within the class of query-output independently noise-adding mechanisms, we derive the optimal noise distribution for (0 , δ ) -differential privacy under a general cost-minimization framework similar to Ghosh et al. (2009); Geng and Viswanath (2016b); Geng et al. (2015); Geng and Viswanath (2016a).”
Differentially Private Uniformly Most Powerful Tests for Binomial Data (2018, methodology): “Related work on developing optimal DP mechanisms for general loss functions such as Geng and Viswanath (2016a) and Ghosh et al. (2009), give mechanisms that optimize symmetric convex loss functions, centered at a real statistic.”
Local Differential Privacy-Based Federated Learning for Internet of Things (2020, background): “[8]–[11] propose a series of staircase mechanisms under centralized differential privacy for optimal utility in terms of the minimum expectation of noise amplitude and power added to the query output considering privacy budgets ǫ → 0 and ǫ → +∞.”
Differential Privacy Techniques for Cyber Physical Systems: A Survey (2018, background): “The branch of differential privacy itself is divided further into two major categories: ( i ) Distribution Optimization [54]–[59], in this branch, the probability density function of differential privacy is optimized without taking in account the dataset.”