Privacy Amplification

Privacy amplification techniques aim to enhance the privacy guarantees of differentially private algorithms by strategically introducing randomness, primarily through subsampling and shuffling. Current research focuses on refining privacy accounting methods for various algorithms like DP-SGD and DP-FTRL, investigating the impact of correlated noise and exploring the interplay between privacy amplification and model architectures like matrix factorization. These advancements are crucial for enabling the responsible use of private data in machine learning, particularly in federated learning and other distributed settings, by improving the privacy-utility trade-off.

Papers

August 19, 2024

Differentially Private Stochastic Gradient Descent with Fixed-Size Minibatches: Tighter RDP Guarantees with or without Replacement
Jeremiah Birrell, Reza Ebrahimi, Rouzbeh Behnia, Jason Pacheco
Private Stochastic Private Stochastic Gradient Descent Subsampling Method Privacy Amplification Dp SGD Independent Minibatching

July 9, 2024

It's Our Loss: No Privacy Amplification for Hidden State DP-SGD With Non-Convex Loss
Meenatchi Sundaram Muthu Selva Annamalai
PRIvacy Leakage Convex Loss Interest Loss Private Stochastic Gradient Descent Privacy Amplification

June 26, 2024

Beyond Statistical Estimation: Differentially Private Individual Computation via Shuffling
Shaowei Wang, Changyu Dong, Xiangfu Song, Jin Li, Zhili Zhou, Di Wang, Han Wu
Differential Privacy Parameter Estimation Shuffle Model Privacy Amplification Privacy Preserving Computation

May 24, 2024

Scaling up the Banded Matrix Factorization Mechanism for Differentially Private ML
Ryan McKenna
Matrix Factorization Privacy Preserving Machine Learning Privacy Amplification Private Machine Learning Noise Optimization

May 23, 2024

Tighter Privacy Auditing of DP-SGD in the Hidden State Threat Model
Tudor Cebere, Aurélien Bellet, Nicolas Papernot
Threat Model Empirical Privacy Privacy Amplification Private Optimization

May 2, 2024

The Privacy Power of Correlated Noise in Decentralized Learning
Youssef Allouah, Anastasia Koloskova, Aymane El Firdoussi, Martin Jaggi, Rachid Guerraoui
Differential Privacy Decentralized Learning Privacy Utility Privacy Amplification Correlated Noise Network Sparsity Decentralized SGD

March 7, 2024

February 16, 2024

TernaryVote: Differentially Private, Communication Efficient, and Byzantine Resilient Distributed Optimization on Heterogeneous Data
Richeng Jin, Yujie Gu, Kai Yue, Xiaofan He, Zhaoyang Zhang, Huaiyu Dai
Differential Privacy Heterogeneous Data Communication Efficiency Byzantine Robust Privacy Amplification Ternary Random Variable

December 30, 2023

ConfusionPrompt: Practical Private Inference for Online Large Language Models
Peihua Mai, Youjia Yang, Ran Yan, Rui Ye, Yan Pang
Large Language Model Privacy Amplification Black Box Large Language Model Private Inference Group Privacy

December 21, 2023

R\'enyi Pufferfish Privacy: General Additive Noise Mechanisms and Privacy Amplification by Iteration
Clément Pierquin, Aurélien Bellet, Marc Tommasi, Matthieu Boussard
Differential Privacy Privacy Amplification Additive Noise Mechanistic Understanding Private Convex Optimization Pufferfish Privacy

December 20, 2023

Enhancing Trade-offs in Privacy, Utility, and Computational Efficiency through MUltistage Sampling Technique (MUST)
Xingyuan Zhao, Ruyu Zhou, Fang Liu
Differential Privacy Privacy Preserving Privacy Policy Task Utility Computational Efficiency Subsampling Method Privacy Amplification

December 14, 2023

Privacy Amplification by Iteration for ADMM with (Strongly) Convex Objective Functions
T-H. Hubert Chan, Hao Xie, Mengshi Zhao
Primal Dual Privacy Amplification ADMM Algorithm Convex Objective Unfolded D ADMM

October 24, 2023

Privacy Amplification for Matrix Mechanisms
Christopher A. Choquette-Choo, Arun Ganesh, Thomas Steinke, Abhradeep Thakurta
Privacy Utility Differential Privacy Guarantee Differential Privacy SGD Privacy Amplification

July 30, 2023

Shuffled Differentially Private Federated Learning for Time Series Data Analytics
Chenxi Huang, Chaoyang Jiang, Zhenghua Chen
Time Series Local Differential Privacy Private Federated Learning Privacy Amplification Time Series Datasets Privacy Preserving Federated Learning

July 5, 2023

Personalized Privacy Amplification via Importance Sampling
Dominik Fay, Sebastian Mair, Jens Sjölund
Differential Privacy Privacy Preserving Importance Sampling Privacy Amplification Privacy Constraint

June 13, 2023

(Amplified) Banded Matrix Factorization: A unified approach to private training
Christopher A. Choquette-Choo, Arun Ganesh, Ryan McKenna, H. Brendan McMahan, Keith Rush, Abhradeep Thakurta, Zheng Xu
Differential Privacy Unified Framework Matrix Factorization Differential Privacy SGD Privacy Amplification Private Model Training

May 29, 2023

Federated Learning of Gboard Language Models with Differential Privacy
Zheng Xu, Yanxiang Zhang, Galen Andrew, Christopher A. Choquette-Choo, Peter Kairouz, H. Brendan McMahan, Jesse Rosenstock, Yuanbo Zhang
Language Model Differential Privacy Privacy Amplification

April 11, 2023

Echo of Neighbors: Privacy Amplification for Personalized Private Federated Learning with Shuffle Model
Yixuan Liu, Suyun Zhao, Li Xiong, Yuhan Liu, Hong Chen
Personalized Federated Learning Thy Neighbor Shuffle Model Privacy Amplification Model Privacy

April 4, 2023

Privacy Amplification via Compression: Achieving the Optimal Privacy-Accuracy-Communication Trade-off in Distributed Mean Estimation
Wei-Ning Chen, Dan Song, Ayfer Ozgur, Peter Kairouz
Differential Privacy Linear Compression Mean Estimation Privacy Amplification Communication Constraint Privacy Accuracy Frequency Estimation

Privacy Amplification

Papers

Differentially Private Stochastic Gradient Descent with Fixed-Size Minibatches: Tighter RDP Guarantees with or without Replacement

It's Our Loss: No Privacy Amplification for Hidden State DP-SGD With Non-Convex Loss

Beyond Statistical Estimation: Differentially Private Individual Computation via Shuffling

Scaling up the Banded Matrix Factorization Mechanism for Differentially Private ML

Tighter Privacy Auditing of DP-SGD in the Hidden State Threat Model

The Privacy Power of Correlated Noise in Decentralized Learning

Privacy Amplification for the Gaussian Mechanism via Bounded Support

Unified Mechanism-Specific Amplification by Subsampling and Group Privacy Amplification

TernaryVote: Differentially Private, Communication Efficient, and Byzantine Resilient Distributed Optimization on Heterogeneous Data

ConfusionPrompt: Practical Private Inference for Online Large Language Models

R\'enyi Pufferfish Privacy: General Additive Noise Mechanisms and Privacy Amplification by Iteration

Enhancing Trade-offs in Privacy, Utility, and Computational Efficiency through MUltistage Sampling Technique (MUST)

Privacy Amplification by Iteration for ADMM with (Strongly) Convex Objective Functions

Privacy Amplification for Matrix Mechanisms

Shuffled Differentially Private Federated Learning for Time Series Data Analytics

Personalized Privacy Amplification via Importance Sampling

(Amplified) Banded Matrix Factorization: A unified approach to private training

Federated Learning of Gboard Language Models with Differential Privacy

Echo of Neighbors: Privacy Amplification for Personalized Private Federated Learning with Shuffle Model

Privacy Amplification via Compression: Achieving the Optimal Privacy-Accuracy-Communication Trade-off in Distributed Mean Estimation