Locally Differentially Private Heavy Hitter Identification

The notion of Local Differential Privacy (LDP) enables users to answer sensitive questions while preserving their privacy. basic LDP frequency oracle protocol the aggregator estimate any value. But when domain input values is large, finding most frequent values, also known as heavy hitters, by estimating frequencies all possible computationally infeasible. In this paper, we propose an for identifying hitters. our proposed protocol, which call Prefix Extending Method (PEM), are divided into groups, with each group reporting a prefix her We analyze how choose optimal parameters and identify two design principles designing protocols high utility. Experiments show that under same privacy guarantee computational cost, PEM has better utility on both synthetic real-world datasets than existing solutions.