Utility-Optimized Local Differential Privacy Mechanisms for Distribution Estimation
Offered By: USENIX via YouTube
Course Description
Overview
Explore a 21-minute conference talk from USENIX Security '19 that introduces Utility-Optimized Local Differential Privacy (ULDP) mechanisms for distribution estimation. Delve into the innovative approach presented by Takao Murakami from AIST, which aims to enhance the utility of personal data statistics while maintaining privacy protection. Learn about the limitations of traditional Local Differential Privacy (LDP) and how ULDP addresses these issues by providing equivalent privacy guarantees only for sensitive data. Discover two proposed ULDP mechanisms: utility-optimized randomized response and utility-optimized RAPPOR. Examine the concept of personalized ULDP with semantic tags, designed to accommodate varying sensitivity levels among users. Gain insights into the theoretical and experimental evidence demonstrating the superior utility of ULDP mechanisms compared to existing LDP approaches, especially when dealing with a high proportion of non-sensitive data. Understand how ULDP can achieve near non-private mechanism utility in low privacy regimes when most data is non-sensitive.
Syllabus
USENIX Security '19 - Utility-Optimized Local Differential Privacy Mechanisms for
Taught by
USENIX
Related Courses
Never Been KIST - Tor’s Congestion Management Blossoms with Kernel-Informed Socket TransportUSENIX via YouTube Eclipse Attacks on Bitcoin’s Peer-to-Peer Network
USENIX via YouTube Control-Flow Bending - On the Effectiveness of Control-Flow Integrity
USENIX via YouTube Protecting Privacy of BLE Device Users
USENIX via YouTube K-Fingerprinting - A Robust Scalable Website Fingerprinting Technique
USENIX via YouTube