1: \begin{abstract}

2: We study spectral density estimation under local differential privacy. A\-no\-ny\-mi\-zation is achieved through truncation followed by Laplace perturbation. We select our estimator from a set of candidate estimators by a penalized contrast criterion. This estimator is shown to attain nearly the same rate of convergence as the best estimator from the candidate set. A key ingredient of the proof are recent results on concentration of quadratic forms in terms of sub-exponential random variables obtained in \cite{goetze2019concentration}. We illustrate our findings in a small simulation study.

3: \end{abstract}