|
Activity Number:
|
212
- GOVT CSpeed 1
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2021 : 1:30 PM to 3:20 PM
|
|
Sponsor:
|
Government Statistics Section
|
|
Abstract #317973
|
|
|
Title:
|
Disclosure Risk from Homogeneity Attack in Differentially Private Frequency Distribution
|
|
Author(s):
|
Fang Liu and Xingyuan Zhao*
|
|
Companies:
|
University of Notre Dame and University of Notre Dame
|
|
Keywords:
|
differential privacy;
privacy loss parameter;
Laplace mechanism;
Gaussian mechanism;
homogeneity attack;
disclosure risk
|
|
Abstract:
|
Differential privacy (DP) is the state-of-the-art concept that provides robust privacy guarantees on released data. In this work, we examine the robustness of sanitized multi-dimensional frequency distributions via DP mechanisms against homogeneity attack (HA). HA allows adversaries to obtain the exact values on the sensitive attributes for their targets without having to identify them from the released data. To that end, we propose measures for disclosure risk from HA and derive closed-form relationships between the privacy loss parameters from DP and the disclosure risk from HA. The availability of the closed-form relationships assists practical understandings of the abstract concept of DP and privacy loss parameters by putting them in the context of a concrete privacy attack and offer a perspective for choosing privacy loss parameters and implementing DP mechanisms for data sanitization and release in practice. We validate the closed-form mathematical relationships in real-life data sets and demonstrate their application in assessment of the disclosure risk due to HA on sanitized data at various privacy loss parameters.
|
Authors who are presenting talks have a * after their name.
