Is Interaction Necessary for Distributed Private Learning?

By Adam D. Smith, Abhradeep Thakurta, Jalaj Upadhyay. 2017

160 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Science Applications, Federated & Distributed Learning

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

In Federated & Distributed Learning

Citations in the corpus

Privacy Amplification via Random Check-Ins (2020, background): ““None” denotes a na¨ıve scheme (with no privacy ampliﬁcation) where each client is used once in any arbitrary order.”
Privacy Amplification by Iteration (2018, neutral): “Noisy SGD [20], [25] Õ (√ d ε2n ) Õ (√ dk ε2n ) ε”
Heavy Hitters and the Structure of Local Privacy (2017, methodology): “In addition to the intrinsic interest of this problem, LDP algorithms for heavyhitters provide important subroutines for solving many other problems, such as median estimation, convex optimization, and clustering [20, 24].”
Encode, Shuffle, Analyze Privacy Revisited: Formalizations and Empirical Evaluation (2020, background): “(There are other natural variants of attributes based on this encoding scheme e.g., in the context of learning algorithms [28], but these are not considered in this paper.)”
Communication Complexity in Locally Private Distribution Estimation and Heavy Hitters (2019, background): “However, interactive schemes are particularly prone to delays, and require more sophisticated implementation (Smith et al., 2017).”
Locally Differentially Private Frequent Itemset Mining (2018, background): “[31] also propose a protocol for the same problem but without interaction.”
Estimating Smooth GLM in Non-interactive Local Differential Privacy Model with Public Unlabeled Data (2019, methodology): “(Smith et al., 2017; Wang et al., 2018; Zheng et al., 2017) and (Wang et al., 2019b) comprehensively studied this problem.”
AttriGuard: A Practical Defense Against Attribute Inference Attacks via Adversarial Machine Learning (2018, background): “Local differential privacy (LDP): LDP [33–40] is a technique based on ε-differential privacy [41] to protect privacy of an individual user’s data record, i.”
Locally Differentially Private Heavy Hitter Identification (2017, background): “For example, mean estimation [11], [27], frequent itemset mining [29], [37], releasing marginals under LDP [10], [30], [41], key-value pair estimation [40], evolving data monitoring [15], [22], and empirical risk minimization [31], [32].”
Locally Private Gaussian Estimation (2018, background): “Network latency as well as server and user liveness constraints compound this difficulty [22].”
Differentially Private Federated Learning: An Information-Theoretic Perspective (2021, background): “Following the de facto standard of differential privacy (DP) in large-scale model fitting (Bassily et al., 2014; 2019; Chaudhuri & Mishra, 2006; Chaudhuri & Monteleoni, 2009; Chaudhuri et al., 2011; Duchi et al., 2013; Jain & Thakurta, 2014; Jain et al., 2012; Smith et al., 2017; Song et al., 2013; ”
Calibrate: Frequency Estimation and Heavy Hitter Identification with Local Differential Privacy via Incorporating Prior Knowledge (2018, unknown): “Local differential privacy (LDP) [27], [5], [20], [22], [30], [3], [12], [13], [34], [29], [23], [8], [24], [14], [6], [28], [32], [35], [1], a privacy protection mechanism based on differential privacy [15], can address both challenges.”