Differential Privacy

Differential privacy (DP) is a rigorous framework for ensuring data privacy in machine learning by adding carefully calibrated noise to model training processes. Current research focuses on improving the accuracy of DP models, particularly for large-scale training, through techniques like adaptive noise allocation, Kalman filtering for noise reduction, and novel gradient processing methods. This active area of research is crucial for enabling the responsible use of sensitive data in various applications, ranging from healthcare and finance to natural language processing and smart grids, while maintaining strong privacy guarantees.

Papers

November 7, 2023

November 2, 2023

Instance-Specific Asymmetric Sensitivity in Differential Privacy
David Durfee
Differential Privacy Private Estimation Bias Variance Sensitivity Aware Unbiased SGG

November 1, 2023

Adversarially Robust Distributed Count Tracking via Partial Differential Privacy
Zhongzheng Xiong, Xiaoyi Zhu, Zengfeng Huang
Differential Privacy Robust Version Object Counting Robust Algorithm Oblivious Adversary

October 31, 2023

October 30, 2023

October 29, 2023

Differentially Private Permutation Tests: Applications to Kernel Methods
Ilmun Kim, Antonin Schrab
Financial Application Differential Privacy Kernel Method Private Data Analysis Kernel Test Private Hypothesis

October 25, 2023

Privately Aligning Language Models with Reinforcement Learning
Fan Wu, Huseyin A. Inan, Arturs Backurs, Varun Chandrasekaran, Janardhan Kulkarni, Robert Sim
Large Language Model Language Model Reinforcement Learning Differential Privacy Model Matching

October 24, 2023

October 23, 2023

Tractable MCMC for Private Learning with Pure and Gaussian Differential Privacy
Yingyu Lin, Yi-An Ma, Yu-Xiang Wang, Rachel Redberg, Zhiqi Bu
Differential Privacy Posterior Sampling Private Learning Tractable Probabilistic Model Approximate Differential Privacy Approximate Sampling

Differential Privacy

Papers

PrivLM-Bench: A Multi-level Privacy Evaluation Benchmark for Language Models

Causal Discovery Under Local Privacy

An Examination of the Alleged Privacy Threats of Confidence-Ranked Reconstruction of Census Microdata

SoK: Memorisation in machine learning

DP-DCAN: Differentially Private Deep Contrastive Autoencoder Network for Single-cell Clustering

Instance-Specific Asymmetric Sensitivity in Differential Privacy

Adversarially Robust Distributed Count Tracking via Partial Differential Privacy

Compression with Exact Error Distribution for Federated Learning

Verification of Neural Networks Local Differential Classification Privacy

Unified Enhancement of Privacy Bounds for Mixture Mechanisms via $f$-Differential Privacy

Privacy-preserving Federated Primal-dual Learning for Non-convex and Non-smooth Problems with Model Sparsification

Privacy-Preserving Federated Learning over Vertically and Horizontally Partitioned Data for Financial Anomaly Detection

Differentially Private Permutation Tests: Applications to Kernel Methods

Privately Aligning Language Models with Reinforcement Learning

Locally Differentially Private Gradient Tracking for Distributed Online Learning over Directed Graphs

On the Inherent Privacy Properties of Discrete Denoising Diffusion Models

Tractable MCMC for Private Learning with Pure and Gaussian Differential Privacy

Mean Estimation Under Heterogeneous Privacy Demands

Conditional Density Estimations from Privacy-Protected Data

PrivImage: Differentially Private Synthetic Image Generation using Diffusion Models with Semantic-Aware Pretraining