Optimization of UD-LDP with statistical prior knowledge

Gnanakumar Thedchanamoorthy, Michael Bewong, Meisam Mohammady, Tanveer Zia, Zahid Islam

Research output: Book chapter/Published conference paperConference paperpeer-review


In the realm of privacy preservation for the crowd-sourced data collection, Local Differential Privacy (LDP) has emerged as a standard for protecting user data. With the existing literature primarily focused on predefined privacy levels for all users, User Driven Local Differential Privacy (UD-LDP) provides a mechanism that allows users to choose the privacy level that best suits them. However, UD-LDP does not account for heterogeneity in Individual Privacy Needs (IPN) distribution, leading to a one-size-fits-all model that may not optimize data utility. This paper evaluates the shortfall in not accommodating diverse IPN distribution within the UD-LDP framework and proposes a method that recommends optimal number of cohorts and their sizes, that is, the Optimal Cohort Configurations (OCC) for different IPN distributions. The findings suggest OCCs for various privacy distribution types, marking a significant step towards a more personalized and efficient privacy-preserving model.
Original languageEnglish
Title of host publicationOptimization of UD-LDP with statistical prior knowledge
PublisherIEEE Xplore
Number of pages6
Publication statusAccepted/In press - 2024


Dive into the research topics of 'Optimization of UD-LDP with statistical prior knowledge'. Together they form a unique fingerprint.

Cite this