The cost of privacy: optimal rates of convergence for parameter estimation with differential privacy
From MaRDI portal
(Redirected from Publication:2054532)
Abstract: Privacy-preserving data analysis is a rising challenge in contemporary statistics, as the privacy guarantees of statistical methods are often achieved at the expense of accuracy. In this paper, we investigate the tradeoff between statistical accuracy and privacy in mean estimation and linear regression, under both the classical low-dimensional and modern high-dimensional settings. A primary focus is to establish minimax optimality for statistical estimation with the -differential privacy constraint. To this end, we find that classical lower bound arguments fail to yield sharp results, and new technical tools are called for. By refining the "tracing adversary" technique for lower bounds in the theoretical computer science literature, we formulate a general lower bound argument for minimax risks with differential privacy constraints, and apply this argument to high-dimensional mean estimation and linear regression problems. We also design computationally efficient algorithms that attain the minimax lower bounds up to a logarithmic factor. In particular, for the high-dimensional linear regression, a novel private iterative hard thresholding pursuit algorithm is proposed, based on a privately truncated version of stochastic gradient descent. The numerical performance of these algorithms is demonstrated by simulation studies and applications to real data containing sensitive information, for which privacy-preserving statistical methods are necessary.
Recommendations
- Privacy-preserving statistical estimation with optimal convergence rates
- Optimal Privacy-Aware Estimation
- Geometrizing rates of convergence under local differential privacy constraints
- Estimation Efficiency Under Privacy Constraints
- Differentially private distributed parameter estimation
- Comparing approximate and probabilistic differential privacy parameters
- Optimal Schemes for Discrete Distribution Estimation Under Locally Differential Privacy
- Efficient estimation under privacy restrictions in the disclosure problem
- Asymptotically Optimal and Private Statistical Estimation
- Concentrated differential privacy: simplifications, extensions, and lower bounds
Cites work
- A statistical framework for differential privacy
- A unified framework for high-dimensional analysis of M-estimators with decomposable regularizers
- Analyze Gauss: optimal bounds for privacy-preserving principal component analysis
- Fast global convergence of gradient methods for high-dimensional statistical recovery
- Fingerprinting codes and the price of approximate differential privacy
- Finite sample differentially private confidence intervals
- Geometrizing rates of convergence under local differential privacy constraints
- High-dimensional statistics. A non-asymptotic viewpoint
- scientific article; zbMATH DE number 5485440 (Why is no real title available?)
- scientific article; zbMATH DE number 1301967 (Why is no real title available?)
- Iterative hard thresholding for compressed sensing
- Minimax Optimal Procedures for Locally Private Estimation
- On minimax estimation of a sparse normal mean vector
- Privacy-preserving statistical estimation with optimal convergence rates
- Regularized \(M\)-estimators with nonconvexity: statistical and algorithmic theory for local optima
- Sparse spatial autoregressions
- Statistical estimation and optimal recovery
- The algorithmic foundations of differential privacy
- Theory of Cryptography
- What can we learn privately?
Cited in
(45)- Local differential privacy: elbow effect in optimal density estimation and adaptation over Besov ellipsoids
- Geometrizing rates of convergence under local differential privacy constraints
- Differentially private high dimensional sparse covariance matrix estimation
- Differentially private precision matrix estimation
- Differentially private SGD with non-smooth losses
- Differentially private distributed parameter estimation
- Constrained forms of statistical minimax: computation, communication, and privacy
- Comment
- Finite sample differentially private confidence intervals
- Privacy-preserving parametric inference: a case for robust statistics
- A novel adaptive differential privacy algorithm for empirical risk minimization
- Model averaging with privacy-preserving
- Optimal Privacy-Aware Estimation
- Private Sampling: A Noiseless Approach for Generating Differentially Private Synthetic Data
- Econometrics with privacy preservation
- Privacy-preserving statistical estimation with optimal convergence rates
- Privacy aware learning
- A Survey of Differentially Private Regression for Clinical and Epidemiological Research
- Privacy-preserving parameter estimation in distributed cases
- Interactive versus noninteractive locally differentially private estimation: two elbows for the quadratic functional
- On robustness and local differential privacy
- Differentially private inference via noisy optimization
- Differentially private confidence intervals for proportions under stratified random sampling
- Edge differentially private estimation in the -model via jittering and method of moments
- Estimation and Inference for High-Dimensional Generalized Linear Models with Knowledge Transfer
- Gaussian differentially private robust mean estimation and inference
- Distributed optimal subsampling for quantile regression with massive data
- General inferential limits under differential and pufferfish privacy
- Majority vote for distributed differentially private sign selection
- Efficiency in local differential privacy
- On the power of private likelihood-ratio tests for goodness-of-fit in frequency tables
- Differentially private projection-depth-based medians
- A Bias-Accuracy-Privacy Trilemma for Statistical Estimation
- Minimax rates of convergence for sliced inverse regression with differential privacy
- Rate-Optimal Rank Aggregation with Private Pairwise Rankings
- Distributed empirical likelihood inference with privacy guarantees
- Differentially private multivariate statistics with an application to contingency table analysis
- A polynomial time, pure differentially private estimator for binary product distributions
- Not all learnable distribution classes are privately learnable
- Private least absolute deviations with heavy-tailed data
- Privacy-preserving parametric inference for spatial autoregressive model
- Versatile differentially private learning for general loss functions
- Personalized differential privacy for ridge regression under output perturbation
- Differentially private learning beyond the classical dimensionality regime
- Online differentially private inference for linear regression model
This page was built for publication: The cost of privacy: optimal rates of convergence for parameter estimation with differential privacy
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2054532)
