A Novel Privacy Disclosure Risk Measure and Optimizing Privacy Preserving Data Publishing Techniques

Abstract

A tremendous amount of individual-level data is generated each day, with a wide variety of uses. This data often contains sensitive information about individuals, which can be disclosed by “adversaries”. Even when direct identifiers such as social security numbers are masked, an adversary may be able to recognize an individual's identity for a data record by looking at the values of quasi-identifiers (QID), known as identity disclosure, or can uncover sensitive attributes (SA) about an individual through attribute disclosure. In data privacy field, multiple disclosure risk measures have been proposed. These share two drawbacks: they do not consider identity and attribute disclosure concurrently, and they make restrictive assumptions on an adversary's knowledge and disclosure target by assuming certain attributes are QIDs and SAs with clear boundary in between. In this study, we present a Flexible Adversary Disclosure Risk (FADR) measure that addresses these limitations, by presenting a single combined metric of identity and attribute disclosure, and considering all scenarios for an adversary’s knowledge and disclosure targets while providing the flexibility to model a specific disclosure preference. In addition, we employ FADR measure to develop our novel “RU Generalization” algorithm that anonymizes a sensitive dataset to be able to publish the data for public access while preserving the privacy of individuals in the dataset. The challenge is to preserve privacy without incurring excessive information loss. Our RU Generalization algorithm is a greedy heuristic algorithm, which aims at minimizing the combination of both disclosure risk and information loss, to obtain an optimized anonymized dataset. We have conducted a set of experiments on a benchmark dataset from 1994 Census database, to evaluate both our FADR measure and RU Generalization algorithm. We have shown the robustness of our FADR measure and the effectiveness of our RU Generalization algorithm by comparing with the benchmark anonymization algorithm.