Heavy Hitter Estimation over Set-Valued Data with Local Differential Privacy

Explore a conference talk from CCS 2016 focusing on heavy hitter estimation over set-valued data with local differential privacy. Delve into the authors' proposed LDPMiner framework, which addresses privacy concerns in data analysis. Learn about local differential privacy, randomized response techniques, and the RAPPOR algorithm. Examine the problem statement, key observations, and the two-phase design of LDPMiner. Analyze experimental evaluations using synthetic datasets with normal and Laplace distributions, as well as real-world datasets. Gain insights into privacy-preserving data analysis techniques for set-valued data and their applications in computer and communications security.

Intro
Outline
Local Differential Privacy (LDP)
Randomized Response (JASA'65)
RAPPOR CCS'14
Succinct Histogram (SH) STOC'15
Set-valued Data & Heavy Hitters
Problem Statement
Simple Solution
Key Observation 1
LDPMiner Design: Two-Phase Framework
Key Observation 2
Experimental Evaluation
Synthetic Dataset: Normal
Synthetic Dataset: Laplace
Real Datasets
Conclusion