Local differential privacy: elbow effect in optimal density estimation and adaptation over Besov ellipsoids (Q2174979)

Let \(X_{1},\dots,X_{n}\) be samples drawn according to some unknown distribution with a probability density function of \(f\). The authors study the problem of nonparametric estimation of the density \(f\) that have access only to privatised views \(Z_{1},\dots,Z_{n}\) of the original data. They consider a non-interactive privacy mechanism creating the privatised data that satisfies the property of local \(\alpha\)-differential privacy. From the authors' summary: ``\dots, we adopt a recent generalisation of classical minimax theory to the framework of local \(\alpha\)-differential privacy and provide a lower bound on the rate of convergence over Besov spaces \(\mathcal{B}_{pq}^{s}\) under mean integrated \(\mathbb{L}^{r}\)-risk. This lower bound is deteriorated compared to the standard setup without privacy, and reveals a twofold elbow effect. In order to fulfill the privacy requirement, we suggest adding suitably scaled Laplace noise to empirical wavelet coefficients. Upper bounds within (at most) a logarithmic factor are derived under the assumption that \(\alpha\) stays bounded as \(n\) increases: A linear but non-adaptive wavelet estimator is shown to attain the lower bound whenever \( p \ge r\) but provides a slower rate of convergence otherwise. An adaptive non-linear wavelet estimator with appropriately chosen smoothing parameters and thresholding is shown to attain the lower bound within a logarithmic factor for all cases.''