Bridging Privacy and Robustness for Trustworthy Machine Learning
Not on the map. Why?
Abstract (arXiv)
The widespread adoption of machine learning necessitates robust privacy protection alongside algorithmic resilience. While Local Differential Privacy (LDP) provides foundational guarantees, sophisticated adversaries with prior knowledge demand more nuanced Bayesian privacy notions, such as Maximum Bayesian Privacy (MBP) and Average Bayesian Privacy (ABP), first introduced by citezhang2022no. Concurrently, machine learning systems require inherent robustness against data perturbations and adversarial manipulations. This paper systematically investigates the intricate theoretical relationships among LDP, MBP, and ABP. Crucially, we bridge these privacy concepts with algorithmic robustness, particularly within the Probably Approximately Correct (PAC) learning framework. Our work demonstrates that privacy-preserving mechanisms inherently confer PAC robustness. We present key theoretical results, including the formalization of the established LDP-MBP relationship, novel bounds between MBP and ABP, and a proof demonstrating PAC robustness from MBP. Furthermore, we establish a novel theoretical relationship quantifying how privacy leakage directly influences an algorithm's input robustness. These results provide a unified theoretical framework for understanding and optimizing the privacy-robustness trade-off, paving the way for the development of more secure, trustworthy, and resilient machine learning systems.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 31 of its 62 references that are in the map. First posted on arXiv 25 Mar 2024.
Builds on
31 of 62 references.
References · 31
- 2014Adam: A Method for Stochastic Optimizationcited by 36,960
- 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,857
- 2016Deep Learning with Differential Privacycited by 2,682
- 2016Membership Inference Attacks against Machine Learning Modelscited by 2,221
- 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,510
- 2019Certified Adversarial Robustness via Randomized Smoothingcited by 1,324
- 2019Deep Leakage from Gradientscited by 1,004
- 2020Inverting Gradients -- How easy is it to break privacy in federated learning?cited by 707
- 2018Certified Robustness to Adversarial Examples with Differential Privacycited by 462
- 2020iDLG: Improved Deep Leakage from Gradientscited by 367
- 2021See through Gradients: Image Batch Recovery via GradInversioncited by 320
- 2017Collecting Telemetry Data Privatelycited by 147
19 more
- 2016Measuring Neural Net Robustness with Constraintscited by 90
- 2021Improved OOD Generalization via Adversarial Training and Pre-trainingcited by 66
- 2019On a generalization of the Jensen-Shannon divergence and the JS-symmetrization of distances relying on abstract meanscited by 61
- 2022Robustness Implies Privacy in Statistical Estimationcited by 47
- 2017The Price of Differential Privacy For Online Learningcited by 44
- 2016Differential Privacy as a Mutual Information Constraintcited by 41
- 2022No Free Lunch Theorem for Security and Utility in Federated Learningcited by 39
- 2016On the Theory and Practice of Privacy-Preserving Bayesian Data Analysiscited by 34
- 2019SoK: Differential Privaciescited by 33
- 2014Sample Complexity Bounds on Differentially Private Learning via Communication Complexitycited by 31
- 2023From Robustness to Privacy and Backcited by 27
- 2022Robustness Implies Generalization via Data-Dependent Generalization Boundscited by 23
- 2013Local Privacy and Minimax Bounds: Sharp Rates for Probability Estimationcited by 19
- 2019A unified view on differential privacy and robustness to adversarial examplescited by 9
- 2020A Computational Separation between Private Learning and Online Learningcited by 8
- 2023Towards Achieving Near-optimal Utility for Privacy-Preserving Federated Learning via Data Generation and Parameter Distortioncited by 6
- 2023Theoretically Principled Federated Learning for Balancing Privacy and Utilitycited by 6
- 2023A Meta-learning Framework for Tuning Parameters of Protection Mechanisms in Trustworthy Federated Learningcited by 6
- 2018Context-aware Data Aggregation with Localized Information Privacycited by 2
Built on by
Citing papers per year: 2024: 1.
1 paper.