Revealing information while preserving privacy

By Irit Dinur, Kobbi Nissim. 2003

1,261 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Safety, Robustness & Interpretability

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 Computational Theory and Mathematics

In Safety, Robustness & Interpretability

Citations in the corpus

The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks (2018, background): “Differential privacy [13,15,16] is a property that an algorithm can satisfy which bounds the information it can leak about its inputs.”
Calibrating Noise to Sensitivity in Private Data Analysis (2006, background): “We continue a line of research initiated by Dinur and Nissim (2003) on privacy in statistical databases.”
Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures (2015, background): “Several authors have explored linear reconstruction attacks [10,12,23], which work as follows: given some released information y and assuming a hidden vector s of sensitive features (e.”
Marginal Release Under Local Differential Privacy (2017, background): “Differential Privacy (DP) [8, 11, 10], unlike its precursor privacy deﬁnitions, provides semantic mathematical promises on individuals’ privacy, interpreted as statistical properties of the output distribution of a randomized algorithm.”
Tight Lower Bounds for Differentially Private Selection (2017, methodology): “Another line of work proves lower bounds in differential privacy via reconstruction attacks [13], [32], [33], [34], [35], [36], [37].”
The Composition Theorem for Differential Privacy (2013, background): “When composing real-valued queries, the Gaussian mechanism is a popular choice [2], [3], [13], [14], [24].”
Boosting and Differential Privacy (2010, methodology): “Moreover, we revisit the “evolution of confidence” arguments due to Dinur, Dwork, and Nissim [5], [12] and show that, for achieving (ε, δ) differential privacy of k-fold composition, the privacy parameter ε degrades as kε(2) + √ kε rather than linearly as kε.”
Comprehensive Privacy Analysis of Deep Learning: Stand-alone and Federated Learning under Passive and Active White-box Inference Attacks (2018, background): “In a reconstruction attack, the attacker’s objective is to infer attributes of the records in the training set [2], [3].”
Concentrated Differential Privacy (2016, background): “The Fundamental Law of Information Recovery states, informally, that “overly accurate” estimates of “too many” statistics completely destroys privacy ([DN03] et sequelae ).”
Privacy Aware Learning (2012, background): “More recently, there has been a large amount of computationally oriented work on privacy [Dwork et al. 2006; Dwork 2008; Zhou et al. 2009a; Wasserman and Zhou 2010; Hall et al. 2011; Dinur and Nissim 2003; Blum et al. 2008; Chaudhuri et al. 2011; Rubinstein et al. 2012].”
Dependence Makes You Vulnberable: Differential Privacy Under Dependent Tuples (2016, background): “The coefficient allows accurate computation of the query sensitivity for dependent data, thus minimizing the noise that needs to be added providing better utility at the same privacy level.”
Differentially Private Empirical Risk Minimization: Efficient Algorithms and Tight Error Bounds (2014, background): “Kasiviswanathan et al. [32] show how the results of many convex ERM problems can be combined to carry out reconstruction attacks in the spirit of Dinur and Nissim [16].”
Algorithmic stability for adaptive data analysis (2015, methodology): “In particular, there is an algorithm (based on [DN03]) that, after seeing the empirical answer to k = O ( α 2 n ) random queries, can ﬁnd a query such that the empirical answer and the correct answer di ﬀ er by α .”
Learning in a Large Function Space: Privacy-Preserving Mechanisms for SVM Learning (2009, background): “Early work on private interactive mechanisms focused on approximating real- and vector-valued statistics (e.g., Dinur and Nissim, 2003; Blum et al., 2005; Dwork et al., 2006; Dwork, 2006; Barak et al., 2007).”
Distributed Private Data Analysis: On Simultaneously Solving How and What (2008, methodology): “However, (to the best of our understanding) these implications of [7, 12] do not imply the lowerbounds we get for local interactive protocols and distributed protocols.”
On the geometry of differential privacy (2009, background): “In the setting looked at by Dinur and Nissim [DN03], the database y ∈ { 0 , 1 } N consists of one private bit for each individual and each query ask for the number of 1’s amongst a (random) subset on [ N ].”
Comprehensive Privacy Analysis of Deep Learning: Passive and Active White-box Inference Attacks against Centralized and Federated Learning (2018, background): “In a reconstruction attack, the attacker’s objective is to infer attributes of the records in the training set [2], [3].”