Renyi Differential Privacy

Abstract

We propose a natural relaxation of differential privacy based on the Renyi divergence. Closely related notions have appeared in several recent papers that analyzed composition of differentially private mechanisms. We argue that the useful analytical tool can be used as a privacy definition, compactly and accurately representing guarantees on the tails of the privacy loss. We demonstrate that the new definition shares many important properties with the standard definition of differential privacy, while additionally allowing tighter analysis of composite heterogeneous mechanisms.

References in corpus (2)

• Deep Learning with Differential Privacy
• Multiple Source Adaptation and the Renyi Divergence

Cited by in corpus (239)

• Deep Learning with Differential Privacy
• Precision Health Data: Requirements, Challenges and Existing Techniques for Data Security and Privacy
• Scalable Private Learning with PATE
• Protection Against Reconstruction and Its Applications in Private Federated Learning
• Federated Learning with Bayesian Differential Privacy
• DarkneTZ: Towards Model Privacy at the Edge using Trusted Execution Environments
• Decentralized Federated Learning through Proxy Model Sharing
• How to DP-fy ML: A Practical Guide to Machine Learning with Differential Privacy
• Federated Learning and Differential Privacy: Software tools analysis, the Sherpa.ai FL framework and methodological guidelines for preserving data privacy
• Differentially Private Synthetic Medical Data Generation using Convolutional GANs
• Privacy Amplification by Iteration
• IBM Federated Learning: an Enterprise Framework White Paper V0.1
• FedDPGAN: Federated Differentially Private Generative Adversarial Networks Framework for the Detection of COVID-19 Pneumonia
• Gaussian Differential Privacy
• When Machine Learning Meets Privacy: A Survey and Outlook
• Adversarial Attacks Against Deep Generative Models on Data: A Survey
• Rényi Differential Privacy of the Sampled Gaussian Mechanism
• A Critical Review on the Use (and Misuse) of Differential Privacy in Machine Learning
• Differentially Private Learning with Adaptive Clipping
• Evaluating Differentially Private Machine Learning in Practice
• Privacy-Preserving Machine Learning: Methods, Challenges and Directions
• Differential Privacy Has Disparate Impact on Model Accuracy
• Large Language Models Can Be Strong Differentially Private Learners
• Differentially Private Learning Needs Better Features (or Much More Data)
• Privately Fine-Tuning Large Language Models with Differential Privacy
• The Federation Strikes Back: A Survey of Federated Learning Privacy Attacks, Defenses, Applications, and Policy Landscape
• Recent Advances of Differential Privacy in Centralized Deep Learning: A Systematic Survey
• Advances in Robust Federated Learning: A Survey with Heterogeneity Considerations
• DP-Forward: Fine-tuning and Inference on Language Models with Differential Privacy in Forward Pass
• Proving Differential Privacy with Shadow Execution
• The Cost of Privacy: Rates of Convergence for Parameter Estimation with Differential Privacy
• DataLens: Scalable Privacy Preserving Training via Gradient Compression and Aggregation
• The OARF Benchmark Suite: Characterization and Implications for Federated Learning Systems
• Federated and Differentially Private Learning for Electronic Health Records
• Privacy and Trust Redefined in Federated Machine Learning
• Generative Models for Effective ML on Private, Decentralized Datasets
• On the Protection of Private Information in Machine Learning Systems: Two Recent Approaches
• Differentially Private Graph Classification with GNNs
• The Skellam Mechanism for Differentially Private Federated Learning
• The Discrete Gaussian for Differential Privacy
• Federated Boosted Decision Trees with Differential Privacy
• Lower Bounds for Locally Private Estimation via Communication Complexity
• GS-WGAN: A Gradient-Sanitized Approach for Learning Differentially Private Generators
• Pointwise Maximal Leakage
• Chorus: a Programming Framework for Building Scalable Differential Privacy Mechanisms
• Not one but many Tradeoffs: Privacy Vs. Utility in Differentially Private Machine Learning
• Privacy-Preserving Distributed Deep Learning for Clinical Data
• Do Not Let Privacy Overbill Utility: Gradient Embedding Perturbation for Private Learning
• Differentially-Private "Draw and Discard" Machine Learning
• When differential privacy meets NLP: The devil is in the detail
• The Distributed Discrete Gaussian Mechanism for Federated Learning with Secure Aggregation
• Guidelines for Implementing and Auditing Differentially Private Systems
• Approximate Span Liftings
• Confidentiality Protection in the 2020 US Census of Population and Housing
• DP-MERF: Differentially Private Mean Embeddings with Random Features for Practical Privacy-Preserving Data Generation
• Opacus: User-Friendly Differential Privacy Library in PyTorch
• Differential privacy for medical deep learning: methods, tradeoffs, and deployment implications
• Practical and Private (Deep) Learning without Sampling or Shuffling
• Improving Deep Learning with Differential Privacy using Gradient Encoding and Denoising
• Average-Case Averages: Private Algorithms for Smooth Sensitivity and Mean Estimation
• Recent advances in deep learning theory
• Membership Inference Attack Susceptibility of Clinical Language Models
• Individual Privacy Accounting via a Renyi Filter
• D2P-Fed: Differentially Private Federated Learning With Efficient Communication
• Differentially Private Vertical Federated Clustering
• A Primer on Private Statistics
• Understanding Unintended Memorization in Federated Learning
• Reviewing and Improving the Gaussian Mechanism for Differential Privacy
• Quantum Pufferfish Privacy: A Flexible Privacy Framework for Quantum Systems
• DP-LSSGD: A Stochastic Optimization Method to Lift the Utility in Privacy-Preserving ERM
• A Differentially Private Framework for Deep Learning with Convexified Loss Functions
• Really Useful Synthetic Data -- A Framework to Evaluate the Quality of Differentially Private Synthetic Data
• DPIS: An Enhanced Mechanism for Differentially Private SGD with Importance Sampling
• Differentially Private Federated Learning for Cancer Prediction
• Training Production Language Models without Memorizing User Data
• Utility-Optimized Local Differential Privacy Mechanisms for Distribution Estimation
• Measuring Data Leakage in Machine-Learning Models with Fisher Information
• Assessing differentially private deep learning with Membership Inference
• Assisted Learning: A Framework for Multi-Organization Learning
• That which we call private
• Information-theoretic metrics for Local Differential Privacy protocols
• Differentially Private Accelerated Optimization Algorithms
• Proximal Langevin Algorithm: Rapid Convergence Under Isoperimetry
• Optimal Accounting of Differential Privacy via Characteristic Function
• Fault Tolerance of Neural Networks in Adversarial Settings
• Privacy and Utility Tradeoff in Approximate Differential Privacy
• The Cost of Privacy in Generalized Linear Models: Algorithms and Minimax Lower Bounds
• RDP-GAN: A Rényi-Differential Privacy based Generative Adversarial Network
• Computing Tight Differential Privacy Guarantees Using FFT
• SoK: Machine Learning Governance
• Efficient Privacy-Preserving Stochastic Nonconvex Optimization
• Concurrent Composition Theorems for Differential Privacy
• Private Post-GAN Boosting
• Toward Evaluating Re-identification Risks in the Local Privacy Model
• Differentially Private Federated Learning with Laplacian Smoothing
• CaPC Learning: Confidential and Private Collaborative Learning
• On the Privacy-Utility Tradeoff in Peer-Review Data Analysis
• How reparametrization trick broke differentially-private text representation learning
• DTGAN: Differential Private Training for Tabular GANs
• Bayesian Differential Privacy for Machine Learning
• Kamino: Constraint-Aware Differentially Private Data Synthesis
• Federated Intrusion Detection for IoT with Heterogeneous Cohort Privacy
• Federated $f$-Differential Privacy
• Element Level Differential Privacy: The Right Granularity of Privacy
• Fast and Memory Efficient Differentially Private-SGD via JL Projections
• Advances in Differential Privacy and Differentially Private Machine Learning
• Fast Dimension Independent Private AdaGrad on Publicly Estimated Subspaces
• Evading Curse of Dimensionality in Unconstrained Private GLMs via Private Gradient Descent
• The Effect of Quantization in Federated Learning: A Rényi Differential Privacy Perspective
• Practical Privacy Filters and Odometers with Rényi Differential Privacy and Applications to Differentially Private Deep Learning
• A(DP)${}^2$SGD: Asynchronous Decentralized Parallel Stochastic Gradient Descent with Differential Privacy
• PRECAD: Privacy-Preserving and Robust Federated Learning via Crypto-Aided Differential Privacy
• On the Convergence and Calibration of Deep Learning with Differential Privacy
• Privacy-preserving Non-negative Matrix Factorization with Outliers
• Locally Differentially Private Data Collection and Analysis
• Neither Private Nor Fair: Impact of Data Imbalance on Utility and Fairness in Differential Privacy
• Gradient Perturbation is Underrated for Differentially Private Convex Optimization
• Dynamic Differential-Privacy Preserving SGD
• Capacity Bounded Differential Privacy
• Obfuscation via Information Density Estimation
• The Complexity of Verifying Boolean Programs as Differentially Private
• Differentially Private GANs for Generating Synthetic Indoor Location Data
• Privacy-preserving Q-Learning with Functional Noise in Continuous State Spaces
• A Fairness Analysis on Private Aggregation of Teacher Ensembles
• Private Non-smooth Empirical Risk Minimization and Stochastic Convex Optimization in Subquadratic Steps
• Tighter Generalization Bounds for Iterative Differentially Private Learning Algorithms
• On the Importance of Difficulty Calibration in Membership Inference Attacks
• Complex-valued Federated Learning with Differential Privacy and MRI Applications
• Hyperparameter Tuning with Renyi Differential Privacy
• Voting-based Approaches For Differentially Private Federated Learning
• Improved Differentially Private Euclidean Distance Approximation
• Renyi Differential Privacy of the Subsampled Shuffle Model in Distributed Learning
• Private and Communication-Efficient Edge Learning: A Sparse Differential Gaussian-Masking Distributed SGD Approach
• Improved Differentially Private Decentralized Source Separation for fMRI Data
• Duet: An Expressive Higher-order Language and Linear Type System for Statically Enforcing Differential Privacy
• Differentially Private Deep Learning with Direct Feedback Alignment
• Differentially Private SGD with Non-Smooth Losses
• Robustness of Maximal $\alpha$-Leakage to Side Information
• Data Sanitisation Protocols for the Privacy Funnel with Differential Privacy Guarantees
• PEARL: Data Synthesis via Private Embeddings and Adversarial Reconstruction Learning
• Decentralized Differentially Private Segmentation with PATE
• Near Instance-Optimality in Differential Privacy
• Machine Unlearning via Algorithmic Stability
• Lossless Compression of Efficient Private Local Randomizers
• Private Multi-Task Learning: Formulation and Applications to Federated Learning
• Stochastic Adaptive Line Search for Differentially Private Optimization
• Exactly Optimal and Communication-Efficient Private Estimation via Block Designs
• Synthetic Tabular Data: Methods, Attacks and Defenses
• Private Language Model Adaptation for Speech Recognition
• Bounding, Concentrating, and Truncating: Unifying Privacy Loss Composition for Data Analytics
• Differential privacy enables fair and accurate AI-based analysis of speech disorders while protecting patient data
• Measuring Information Leakage in Non-stochastic Brute-Force Guessing
• P3GM: Private High-Dimensional Data Release via Privacy Preserving Phased Generative Model
• Super-convergence and Differential Privacy: Training faster with better privacy guarantees
• Robust and Private Learning of Halfspaces
• A unified view on differential privacy and robustness to adversarial examples
• The Trade-Offs of Private Prediction
• Towards practical differentially private causal graph discovery
• Robustness, Privacy, and Generalization of Adversarial Training
• Effective and Privacy preserving Tabular Data Synthesizing
• DP-REC: Private & Communication-Efficient Federated Learning
• Evaluating Differential Privacy on Correlated Datasets Using Pointwise Maximal Leakage
• PrivateXR: Defending Privacy Attacks in Extended Reality Through Explainable AI-Guided Differential Privacy
• Quantum Local Differential Privacy and Quantum Statistical Query Model
• Node-Level Differentially Private Graph Neural Networks
• Releasing Graph Neural Networks with Differential Privacy Guarantees
• DPlis: Boosting Utility of Differentially Private Deep Learning via Randomized Smoothing
• Differential Privacy Dynamics of Langevin Diffusion and Noisy Gradient Descent
• Computing Differential Privacy Guarantees for Heterogeneous Compositions Using FFT
• DPD-InfoGAN: Differentially Private Distributed InfoGAN
• Tight Differential Privacy for Discrete-Valued Mechanisms and for the Subsampled Gaussian Mechanism Using FFT
• Upper Bounds on the Generalization Error of Private Algorithms for Discrete Data
• Sharp Composition Bounds for Gaussian Differential Privacy via Edgeworth Expansion
• Linear and Range Counting under Metric-based Local Differential Privacy
• Differential privacy with partial knowledge
• Model-based Differentially Private Data Synthesis and Statistical Inference in Multiply Synthetic Differentially Private Data
• Don't Generate Me: Training Differentially Private Generative Models with Sinkhorn Divergence
• Private Stochastic Convex Optimization: Efficient Algorithms for Non-smooth Objectives
• Noise Variance Optimization in Differential Privacy: A Game-Theoretic Approach Through Per-Instance Differential Privacy
• Optimizing Fitness-For-Use of Differentially Private Linear Queries
• Perceptual Indistinguishability-Net (PI-Net): Facial Image Obfuscation with Manipulable Semantics
• Generating private data with user customization
• Faster Differentially Private Samplers via Rényi Divergence Analysis of Discretized Langevin MCMC
• Enhancing the Antidote: Improved Pointwise Certifications against Poisoning Attacks
• Sensitivity analysis in differentially private machine learning using hybrid automatic differentiation
• Utility/Privacy Trade-off through the lens of Optimal Transport
• Privacy Preserving Off-Policy Evaluation
• FedPower: Privacy-Preserving Distributed Eigenspace Estimation
• AirMixML: Over-the-Air Data Mixup for Inherently Privacy-Preserving Edge Machine Learning
• Photonic Differential Privacy with Direct Feedback Alignment
• Privacy-Aware Rejection Sampling
• Strong Asymptotic Composition Theorems for Mutual Information Measures
• Muffliato: Peer-to-Peer Privacy Amplification for Decentralized Optimization and Averaging
• Divergences on Monads for Relational Program Logics
• High-Dimensional Private Empirical Risk Minimization by Greedy Coordinate Descent
• From Noisy Fixed-Point Iterations to Private ADMM for Centralized and Federated Learning
• Location Trace Privacy Under Conditional Priors
• Data Poisoning and Leakage Analysis in Federated Learning
• Differential Privacy, Linguistic Fairness, and Training Data Influence: Impossibility and Possibility Theorems for Multilingual Language Models
• An Adaptive and Fast Convergent Approach to Differentially Private Deep Learning
• Understanding the Interplay between Privacy and Robustness in Federated Learning
• Oneshot Differentially Private Top-k Selection
• Improving the Algorithm of Deep Learning with Differential Privacy
• Free Gap Information from the Differentially Private Sparse Vector and Noisy Max Mechanisms
• Learning Numeric Optimal Differentially Private Truncated Additive Mechanisms
• TableGAN-MCA: Evaluating Membership Collisions of GAN-Synthesized Tabular Data Releasing
• Privacy-Preserving Distributed Processing: Metrics, Bounds, and Algorithms
• Privacy-preserving Data Sharing on Vertically Partitioned Data
• Quantifying Membership Privacy via Information Leakage
• Private Knowledge Transfer via Model Distillation with Generative Adversarial Networks
• A Privacy-Preserving and Trustable Multi-agent Learning Framework
• Towards Assessment of Randomized Smoothing Mechanisms for Certifying Adversarial Robustness
• Amplifying Rényi Differential Privacy via Shuffling
• Privacy Amplification via Iteration for Shuffled and Online PNSGD
• A Graph Symmetrisation Bound on Channel Information Leakage under Blowfish Privacy
• Splintering with distributions: A stochastic decoy scheme for private computation
• Partial sensitivity analysis in differential privacy
• Renyi Differentially Private ADMM for Non-Smooth Regularized Optimization
• Leave-one-out Unfairness
• DuetSGX: Differential Privacy with Secure Hardware
• Concurrent Composition of Differential Privacy
• Simulation of Random Variables under Rényi Divergence Measures of All Orders
• NeuralDP Differentially private neural networks by design
• Topology-aware Differential Privacy for Decentralized Image Classification
• Towards General-purpose Infrastructure for Protecting Scientific Data Under Study
• Differentially Private ADMM Algorithms for Machine Learning
• Differentially Private Multivariate Statistics with an Application to Contingency Table Analysis
• Selective Differential Privacy for Language Modeling
• Privacy with Estimation Guarantees
• Improving Differentially Private SGD via Randomly Sparsified Gradients
• DP-CDA: An Algorithm for Enhanced Privacy Preservation in Dataset Synthesis Through Randomized Mixing
• BGTplanner: Maximizing Training Accuracy for Differentially Private Federated Recommenders via Strategic Privacy Budget Allocation
• Controlled privacy leakage propagation throughout overlapping grouped learning
• MCMC for Bayesian estimation of Differential Privacy from Membership Inference Attacks
• A Novel Approach to Differential Privacy with Alpha Divergence
• Robust and Differentially Private Principal Component Analysis
• Cyclic Adaptive Private Synthesis for Sharing Real-World Data in Education
• Smoothed Differential Privacy
• Tunable Measures for Information Leakage and Applications to Privacy-Utility Tradeoffs