PRIVACY CHARACTERIZATION AND QUANTIFICATION IN DATA PUBLISHING AN INTELLIGENT FRAMEWORK FOR MEASURING PRIVACY LEAKAGE IN PPDP TECHNIQUES

Abstract

The increasing interest in publishing large-scale individual datasets for medical research, market analysis, and economic studies has raised serious privacy concerns. While numerous Privacy-Preserving Data Publishing (PPDP) techniques such as k-anonymity, l-diversity, and t-closeness have been proposed, most lack a comprehensive and quantifiable privacy characterization model. This paper presents a novel multi-variable privacy characterization and quantification framework that models attributes as a multi-dimensional privacy risk space. The framework redefines prior and posterior adversarial beliefs and analyzes the sensitivity arising from attribute combinations. We demonstrate that privacy leakage cannot be accurately measured using a single metric and therefore propose two new quantification metrics: Distribution Leakage and Entropy Leakage. Using these metrics, we systematically evaluate well-known PPDP techniques. Experimental results show that existing schemes suffer from significant limitations in privacy protection. The proposed framework provides a solid foundation for designing, analyzing, and comparing future privacypreserving data publishing mechanisms, enabling better trade-offs between privacy and data utility.