Explore a 23-minute conference talk on constructing marginal tables from multi-dimensional user data while adhering to Local Differential Privacy (LDP). Delve into the CALM (Consistent Adaptive Local Marginal) protocol, which addresses privacy concerns without relying on trusted third parties. Learn about data collection methods, LDP deployment and applications, and various techniques including Random Response, Strawman Method 2 (AM), and Fourier Transformation Method (FT). Discover how CALM ensures consistency between noisy marginals, constructs k-way marginals, and selects appropriate marginal sets. Examine experimental results on binary and non-binary datasets, assessing performance through Sum of Squared Errors (SSE) and classification accuracy.
CALM - Consistent Adaptive Local Marginal for Marginal Release under Local Differential Privacy
Association for Computing Machinery (ACM) via YouTube
Overview
Syllabus
Intro
Data Collection
Local Differential Privacy (LDP)
Random Response
Deployment of LDP
Application of LDP
Marginal Table
Strawman Method 2 (AM)
Fourier Transformation Method (FT)
Protocol Overview (CALM)
How to consist between noisy marginals (step 3)
How to construct all k-way marginals (step 4)
How to choose a set of marginals (step 1)
Experimental Setup
SSE on Binary Dataset
SSE on Non-binary Dataset
Classification Performance
Conclusion
Taught by
Association for Computing Machinery (ACM)
