Tight Bounds for Differentially Private Anonymized Histograms
share
In this note, we consider the problem of differentially privately (DP) computing an anonymized histogram, which is defined as the multiset of counts of the input dataset (without bucket labels). In the low-privacy regime ϵ≥ 1, we give an ϵ-DP algorithm with an expected ℓ_1-error bound of O(√(n) / e^ϵ). In the high-privacy regime ϵ < 1, we give an Ω(√(n log(1/ϵ) / ϵ)) lower bound on the expected ℓ_1 error. In both cases, our bounds asymptotically match the previously known lower/upper bounds due to [Suresh, NeurIPS 2019].
READ FULL TEXT