Private Learning Algorithm

Private learning algorithms aim to train machine learning models on sensitive data while guaranteeing individual privacy through techniques like differential privacy. Current research focuses on improving the efficiency and accuracy of these algorithms, exploring methods such as leveraging public data, optimizing noise injection (including correlated noise), and refining model architectures (e.g., through feature selection and architecture search) to mitigate the utility loss inherent in privacy-preserving training. This field is crucial for enabling responsible use of sensitive data in machine learning, with applications ranging from healthcare to finance, and ongoing work addresses challenges in both theoretical guarantees and practical implementation.

Papers

October 22, 2024

Privacy-Computation trade-offs in Private Repetition and Metaselection
Kunal Talwar
Private Algorithm Exercise Repetition Private Learning Algorithm Private Hyperparameter Self Selection

February 20, 2024

Revisiting Differentially Private Hyper-parameter Tuning
Zihang Xiang, Tianhao Wang, Chenglong Wang, Di Wang
Differential Privacy Hyper Parameter Tuning Private Learning Algorithm Private Selection Private Hyperparameter

February 13, 2024

Oracle-Efficient Differentially Private Learning with Public Data
Adam Block, Mark Bun, Rathin Desai, Abhishek Shetty, Steven Wu
Differential Privacy Available Datasets Data Distribution Private Learning Privacy Constraint Private Learning Algorithm

October 10, 2023

Correlated Noise Provably Beats Independent Noise for Differentially Private Learning
Christopher A. Choquette-Choo, Krishnamurthy Dvijotham, Krishna Pillutla, Arun Ganesh, Thomas Steinke, Abhradeep Thakurta
Private Learning Additive Noise Correlated Noise Correlation Function Private Learning Algorithm

June 2, 2023

Harnessing large-language models to generate private synthetic text
Alexey Kurakin, Natalia Ponomareva, Umar Syed, Liam MacDermed, Andreas Terzis
Synthetic Data Synthetic Dataset Private Learning Algorithm Private INFORMATION RETRIEVAL

May 28, 2023

Training Private Models That Know What They Don't Know
Stephan Rabanser, Anvith Thudi, Abhradeep Thakurta, Krishnamurthy Dvijotham, Nicolas Papernot
Differential Privacy Selective Prediction Non Private Differential Privacy Constraint Private Learning Algorithm

May 13, 2022

On the Importance of Architecture and Feature Selection in Differentially Private Machine Learning
Wenxuan Bao, Luke A. Bauer, Vincent Bindschaedler
Differential Privacy Feature Selection Privacy Preserving Importance Aware Architecture Design Private Machine Learning Private Learning Algorithm

May 6, 2022

Large Scale Transfer Learning for Differentially Private Image Classification
Harsh Mehta, Abhradeep Thakurta, Alexey Kurakin, Ashok Cutkosky
Differential Privacy Private Stochastic Gradient Descent Privacy Preserving Image Differential Privacy SGD Non Private Private Learning Algorithm

February 24, 2022

Debugging Differential Privacy: A Case Study for Privacy Auditing
Florian Tramer, Andreas Terzis, Thomas Steinke, Shuang Song, Matthew Jagielski, Nicholas Carlini
Differential Privacy Case Study Privacy Guarantee Private Algorithm Differential Privacy Guarantee Empirical Privacy Private Learning Algorithm

November 9, 2021

DP-REC: Private & Communication-Efficient Federated Learning
Aleksei Triastcyn, Matthias Reisser, Christos Louizos
Differential Privacy Privacy Guarantee Compression Technique Entropy Coding Private Learning Algorithm