Publications
カンファレンス (国内) Optimal Variance and Covariance Estimation under Differential Privacy in the Add-Remove Model
Shokichi Takakura, Seng Pei Liew, Satoshi Hasegawa
第18回データ工学と情報マネジメントに関するフォーラム（第24回日本データベース学会年次大会） (DEIM 2026)
2026.3.4
In this paper, we study the problem of estimat- ing the variance and covariance of datasets under differential privacy in the add-remove model. While estimation in the swap model has been extensively studied in the literature, the add-remove model remains less explored and more challenging, as the dataset size must also be kept private. To address this issue, we develop efficient mechanisms for variance and covariance estimation based on the Bézier mechanism, a novel moment-release frame- work that leverages Bernstein bases. We prove that our proposed mechanisms are minimax optimal in the high-privacy regime by estab- lishing new minimax lower bounds. More- over, beyond worst-case scenarios, we analyze instance-wise utility and show that the Bézier- based estimator consistently achieves better utility compared to alternative mechanisms. Finally, we demonstrate the effectiveness of the Bézier mechanism beyond variance and covariance estimation, showcasing its applica- bility to other statistical tasks
Paper :
Optimal Variance and Covariance Estimation under Differential Privacy in the Add-Remove Model
（外部サイト）