Personalized Publishing of Data with Multiple Sensitive Attributes based on Sensitivity Level

Authors

  • Ram Prasad Reddy Sadi, Pandu Ranga Vital Terlapu

Abstract

Data publishing scenario without compromising privacy and utility is challenging and essential for
individuals, researchers and data providers. Much of the research work in this direction assumes that each
individual is associated with only one record in a dataset and has only sensitive attribute, which really is
not realistic in real world scenarios. If the person possessing multiple sensitive attributes appears more
than once in the dataset, several privacy breaches might take place. The practical scenarios in privacy
preserving data publishing with each individual appearing multiple times and each occurrence having
multiple sensitive attributes has not attracted much attention of researchers. We call such datasets as
(1:M:N)-datasets. This paper attempts to provide a new privacy model, (k,l,s)-covering model that tends to
exposure chances in (1:M:N)-dataset distributing.This paper also includes personalization where a user
has the privilege to specify whether to disclose the data or not. We also present an effective generalization
algorithm, (1:M:N)-generalization, as part of the model, to retain privacy and the same time provide utility
for the published data. The model is tested on real world datasets and the results showed excellent
improvement with respect to utility of the data and execution time when compared to other existing
approaches

Published

2020-10-17

Issue

Section

Articles