Differentially Private Heavy Hitter Detection using Federated Analytics

This study aims to improve the performance of prefix-tree-based algorithms for differentially private heavy hitter detection. To enhance this, an adaptive hyperparameter tuning algorithm is introduced. It helps to increase the performance of the system while satisfying computational, communication, and privacy constraints. Different data-selection schemes are explored, and the effects of introducing deny lists during multiple runs of the algorithm are investigated.

Publication date: July 24, 2023
Project Page: N/A
Paper: https://arxiv.org/pdf/2307.11749.pdf