DP-SGD provides privacy guarantees for all data points, but we propose output-specific DP to better account for individual examples.
problem Accounting for individual privacy guarantees in DP-SGD.
method Output-specific ( ε , δ ) (\varepsilon,δ) ( ε , δ ) -DP and an efficient algorithm to investigate individual privacy across datasets. result Most examples enjoy stronger privacy guarantees than the worst-case bound, and there is a correlation between training loss and privacy parameter.
We present an approach for designing correct-by-construction neural networks (and other machine learning models) that are guaranteed to be consistent with a collection of input-output specifications before, during, and after algorithm training. Our method involves designing a constrained predictor for each set of compa…
Machine learning (ML) models trained by differentially private stochastic gradient descent (DP-SGD) have much lower utility than the non-private ones. To mitigate this degradation, we propose a DP Laplacian smoothing SGD (DP-LSSGD) to train ML models with differential privacy (DP) guarantees. At the core of DP-LSSGD is…
New DP-CD method outperforms DP-SGD in solving composite DP-ERM problems.
problem Privacy-preserving machine learning with differential privacy.
method Differentially Private proximal Coordinate Descent (DP-CD) for composite Empirical Risk Minimization (ERM).
result DP-CD outperforms DP-SGD due to larger step sizes and better gradient exploitation.
DOPPLER optimizes DP training with low-pass filtering, improving model accuracy.
problem Privacy concerns in deep learning models and performance degradation of DP optimizers.
method Developed DOPPLER, a low-pass filter for DP optimizers, to reduce privacy noise and enhance model quality.
result DOPPLER optimizers outperform non-DOPPLER counterparts by 3%-10% in test accuracy.
DP-BNNs improve accuracy, privacy, and reliability in neural networks.
problem Balancing privacy and accuracy in neural networks.
method Proposed three DP-BNNs: DP-SGLD, DP-BBP, and DP-MC Dropout.
result DP-SGLD achieves high accuracy under strong privacy guarantees.
Proactive DP optimizes privacy and utility in DP-SGD with a fixed privacy budget.
problem Balancing privacy and utility in differential privacy for machine learning.
method Proposes a pro-active DP framework that allows a-priori selection of DP-SGD parameters to maximize test accuracy.
result Proactive DP can optimize utility of DP-SGD with a fixed privacy budget (ε, δ).
Framework purifies approximate differential privacy to pure differential privacy.
problem Achieving pure differential privacy from approximate differential privacy.
method Randomized post-processing with calibrated noise to eliminate δ parameter.
result First statistically and computationally efficient reduction from approximate DP to pure DP.
Paper defines new topological invariants for DP tangles.
problem Classifying and understanding doubly periodic tangles.
method Organized components into interlinked compounds; introduced axis-motif.
result Directional type is an invariant of DP tangles.
Three DP variants linked, improving SGD privacy bounds.
problem Relating different DP variants for tighter privacy bounds.
method Developed machinery to relate approximate DP to RDP and hypothesis test DP.
result Improved privacy guarantees for noisy SGD.
This paper improves privacy accounting in decentralized FL using f-Differential Privacy.
problem Challenges in accurately quantifying privacy budget in decentralized FL.
method Develops two new f-DP-based accounting methods for decentralized FL.
result Yields tighter (ε,δ) bounds and improved utility compared to existing methods.
New divergences help audit DP in high dimensions.
problem Challenges in auditing DP in high-dimensional data.
method Propose kernel Rényi divergence and its regularized version for auditing.
result Regularized kernel Rényi divergence can be estimated from samples in high dimensions.
DP-Net uses dynamic programming for efficient deep neural network compression.
problem Efficiently compressing deep neural networks while maintaining accuracy.
method Dynamic Programming for optimal weight quantization and clustering-friendly training.
result Achieves up to 77X compression ratio on Wide ResNet with minimal accuracy loss.
Balls-and-Bins sampling improves DP-SGD privacy and utility.
problem Improving privacy and utility in DP-SGD implementations.
method Introducing Balls-and-Bins sampling as an alternative to shuffling in DP-SGD.
result Balls-and-Bins sampling achieves utility comparable to shuffling while offering better privacy amplification.
This paper improves privacy bounds for DP algorithms using f f f -DP.
problem Difficulty in analyzing randomness in DP algorithms due to mixture distributions.
method Derives a closed-form expression for trade-off functions and analyzes f f f -DP. result Enhances privacy of DP-GD with random initialization and shuffling models.
New DP bootstrap method for statistical inference with improved privacy and accuracy.
problem Lack of general techniques for conducting statistical inference under differential privacy.
method DP bootstrap procedure to infer sampling distribution and construct confidence intervals.
result DP bootstrap estimates provide consistent point estimates and asymptotically valid standard CIs.
A new differentiable model for sampling DAGs that speeds up optimization.
problem Efficiently sampling and learning DAG structures in continuous optimization.
method DP-DAG model with VI-DP-DAG for DAG learning from data.
result VI-DP-DAG outperforms other methods in DAG structure and causal mechanism learning.
A new DP algorithm improves privacy in hashing and sampling for search and learning.
problem Improving privacy in hashing and sampling for large-scale applications.
method Combines differential privacy with one permutation hashing and bin-wise consistent weighted sampling.
result Proposes DP-OPH and DP-BCWS algorithms that enhance privacy while maintaining utility.
Gaussian DP improves reporting of ML algorithms' differential privacy guarantees.
problem Incomplete and misleading DP guarantees for ML algorithms.
method Using non-asymptotic Gaussian Differential Privacy (GDP) to provide accurate bounds on privacy profiles of ML algorithms.
result GDP captures the entire privacy profile of DP-SGD and related algorithms with virtually no error.
The paper introduces DP algorithms using random projections and sign random projections for improved privacy in machine learning.
problem Improving differential privacy in machine learning applications.
method Developed algorithms based on random projections and sign random projections, focusing on individual differential privacy (iDP) and standard differential privacy (DP).
result DP-SignOPORP and iDP-SignRP achieve superior performance in differential privacy, especially for small epsilon values.
Recent developments in differentially private (DP) machine learning and DP Bayesian learning have enabled learning under strong privacy guarantees for the training data subjects. In this paper, we further extend the applicability of DP Bayesian learning by presenting the first general DP Markov chain Monte Carlo (MCMC)…
Paper constructs motifs from planar tilings for DP weaves and polycatenanes.
problem Creating complex entangled structures from periodic tilings.
method Combinatorial methodology using polygonal link transformations.
result Predicting the type of motif from a given tiling and polygonal link method.
Lower bounds on query complexity for reconstructing private learner's training data.
problem Query complexity of reconstructing private learner's training data.
method Minimax analysis, Rényi DP, Metric DP framework.
result First known lower bounds on adversary's query complexity for various DP learners.
Paper defines and analyzes mathematical equivalence of periodic tangles.
problem Understanding the equivalence of periodic tangles.
method Established mathematical framework, characterized isotopies, generalized results.
result Characterization of DP tangle equivalence based on motifs.
Improved regret bounds for DP-KLUCB and DP-IMED in Bernoulli bandits.
problem Minimizing regret in stochastic bandits under ε-global Differential Privacy.
method Developed DP versions of KLUCB and IMED, proving tighter lower bounds and matching upper bounds.
result DP-KLUCB and DP-IMED achieve asymptotically optimal regret under ε-global DP.
Differential privacy (DP) is a popular mechanism for training machine learning models with bounded leakage about the presence of specific points in the training data. The cost of differential privacy is a reduction in the model's accuracy. We demonstrate that in the neural networks trained using differentially private …
Framework evaluates privacy cost of non-private pre-processing in DP pipelines.
problem Privacy cost of non-private data-dependent pre-processing in DP machine learning pipelines.
method Establishes upper bounds on overall privacy guarantees using Smooth DP and bounded sensitivity.
result Explicit overall privacy guarantees for various pre-processing algorithms.
New DP mechanisms improve ML privacy-utility-computational tradeoffs.
problem Improving privacy in machine learning with multiple passes over data.
method Formalized DP for adaptive streams, extended matrix factorization techniques, Fourier-transform-based mechanism.
result Substantial improvements in privacy-utility-computational tradeoffs over previous methods.
New DP framework using data truncation for efficient estimation.
problem Differential privacy in unbounded data support.
method Data truncation, exponential family distributions, maximum likelihood estimation, DP stochastic gradient descent.
result Near-optimal sample complexity for Gaussian mean and covariance estimation.
New DP algorithms with margin guarantees for various hypothesis sets.
problem Differential privacy in machine learning with margin guarantees.
method Developed pure and efficient DP learning algorithms for linear, kernel-based, and neural network hypotheses.
result Margin guarantees are independent of input dimension and hypothesis type.
This guide simplifies applying differential privacy to machine learning models.
problem Limited practical guidance for achieving good privacy-utility-computations in ML models.
method Comprehensive self-contained guide covering theory and practical implementation.
result Achieves best possible DP ML model with rigorous privacy guarantees.
This paper provides a method for noise-calibrated inference from DP synthetic data.
problem Inference from DP synthetic data is often miscalibrated and lacks principled uncertainty quantification.
method Release DP sufficient statistics, perform noise-calibrated likelihood-based inference, and optional synthetic data generation.
result Asymptotic normality and valid confidence intervals for the plug-in DP MLE.
Improved DP-SGD on large models achieves high accuracy on image classification tasks.
problem Differentially private image classification often degrades performance.
method Careful hyper-parameter tuning and signal propagation techniques.
result Achieved 81.4% top-1 accuracy on CIFAR-10 under (8, 10^{-5})-DP.
Differential privacy has seen remarkable success as a rigorous and practical formalization of data privacy in the past decade. This privacy definition and its divergence based relaxations, however, have several acknowledged weaknesses, either in handling composition of private algorithms or in analyzing important primi…
Unified bounds for DP risks reduce noise and improve accuracy.
problem Difficult interpretation and calibration of DP mechanisms.
method Hypothesis-testing interpretation of DP ( f f f -DP) and unified bounds. result Unified bounds are tighter and tunable for specific risks.
This paper improves privacy in federated learning without a trusted server.
problem Privacy in federated learning with silos that distrust each other.
method Introduces Inter-Silo Record-Level Differential Privacy (ISRL-DP) and accelerated algorithms for convex and smooth losses.
result Achieves optimal privacy and accuracy tradeoffs in federated learning.
DPConvCNP learns to predict private data accurately and privately.
problem Balancing privacy and accuracy in machine learning models.
method Meta-learning combined with improved DP mechanism.
result DPConvCNP outperforms DP GP baseline, especially on non-Gaussian data.
New method compares DP mechanisms beyond ( ε , δ ) (\varepsilon, δ) ( ε , δ ) pairs.
problem Current DP mechanism comparisons overlook substantial differences.
method Introduces Δ Δ Δ -divergence for robust comparison. result Demonstrates gaps in current DP-SGD practices.
DP-REC combines privacy and communication efficiency in federated learning.
problem Combining privacy and communication efficiency in federated learning.
method DP-REC uses Relative Entropy Coding (REC) for compression and a minor modification for differential privacy.
result DP-REC reduces communication costs while maintaining privacy comparable to state-of-the-art methods.
This work improves privacy-generalization bounds for DP-SGD.
problem Understanding the trade-off between privacy and generalization in machine learning.
method Proved a linear max-information bound for DP-SGD, derived PAC-Bayes and generalization bounds.
result Explicit and controlled complexity terms for DP-SGD-trained models.
Paper proposes DP-SGD and DP-NSGD for differentially private non-convex optimization.
problem Mitigating privacy risks in large model learning.
method Clip or normalize per-sample gradients and add noise for differential privacy.
result Achieved convergence rate of gradient norm for non-convex optimization.
Many applications of machine learning, for example in health care, would benefit from methods that can guarantee privacy of data subjects. Differential privacy (DP) has become established as a standard for protecting learning results. The standard DP algorithms require a single trusted party to have access to the entir…
DP-SGD analysis shows many datapoints leak less privacy than previously thought.
problem Empirical evidence suggests DP-SGD leaks less privacy than current analysis predicts.
method Developed a per-instance DP analysis of DP-SGD, introducing dependence on dataset distribution.
result Formally shows DP-SGD leaks significantly less privacy for many datapoints on common benchmarks.
Improved DP-SGD for variational inference reduces noise and variance.
problem Poor convergence and high variance in variational parameter outputs due to gradient noise in DP-SGD.
method Introduced aligned gradients and iterate averaging to reduce DP-induced noise, and noise-aware posteriors.
result Less noisy gradient estimator and improved parameter estimates for variational inference.
Large batch training with DP-SGD reduces model performance due to implicit bias.
problem Large batch training with DP-SGD reduces model performance.
method The study analyzes the phenomenon of implicit bias in Noisy-SGD (DP-SGD without clipping) and its theoretical solutions for linear models.
result The implicit bias in large batch training with DP-SGD is amplified by additional noise, similar to SGD.
New method improves optimization and DP in FL.
problem Combining strong DP and optimization in FL.
method Combining clipping, momentum, and error feedback.
result Optimal convergence rate and near optimal DP guarantees.
DP-RandP improves privacy-utility tradeoff in DP-SGD by learning priors from random processes.
problem Improving the performance of differentially private stochastic gradient descent (DP-SGD) on private data.
method A three-phase approach that learns priors from images generated by random processes and transfers these priors to private data.
result New state-of-the-art accuracy on CIFAR10, CIFAR100, MedMNIST, and ImageNet for various privacy budgets.
Improved DP algorithms for non-convex optimization with tighter generalization bounds.
problem Private stochastic non-convex optimization in high-dimensional spaces.
method Differential privacy techniques, including adaptive algorithms like DP RMSProp and DP Adam, combined with adaptive data analysis.
result Achieved a sharper rate of p 4 / n \sqrt[4]{p}/\sqrt{n} 4 p / n for population loss, improving upon previous bounds.