Publishing set-valued data via differential privacy

Abstract

Set-valued data provides enormous opportunities for various data mining tasks. In this paper, we study the problem of publishing set-valued data for data mining tasks under the rigorous differential privacy model. All existing data publishing methods for set-valued data are based on partition-based privacy models, for example k -anonymity, which are vulnerable to privacy attacks based on background knowledge. In contrast, differential privacy provides strong privacy guarantees independent of an adversary's background knowledge and computational power. Existing data publishing approaches for differential privacy, however, are not adequate in terms of both utility and scalability in the context of set-valued data due to its high dimensionality. We demonstrate that set-valued data could be efficiently released under differential privacy with guaranteed utility with the help of context-free taxonomy trees. We propose a probabilistic top-down partitioning algorithm to generate a differentially private release, which scales linearly with the input data size. We also discuss the applicability of our idea to the context of relational data. We prove that our result is (∈, δ)-useful for the class of counting queries, the foundation of many data mining tasks. We show that our approach maintains high utility for counting queries and frequent itemset mining and scales to large datasets through extensive experiments on real-life set-valued datasets.

Group privacy-aware disclosure of association graph data

“…1: Graph buyers in a given neighborhood represented by a zipcode). While existing mechanisms [7], [10], [13], [19], [20], [26], [27], [31] have focused on protecting individual's sensitive values, this paper proposes a privacy-preserving data publishing mechanism addressing group privacy concerns when aggregate information about groups of individuals is sensitive and needs protection. In a drug purchase association graph, one may need to protect group privacy at different protection levels depending on the access privilege of the data users.…”

“…Based on the concept of differential privacy [10], there had been many work focused on publishing sensitive datasets through differential privacy constraints [7], [13], [28]. Differential privacy had also been applied to protecting sensitive information in graph datasets such that the released information does not reveal the presence of a sensitive element [9], [17], [25].…”

“…For instance, medical scientists may want to study the outbreaks of new diseases based on the type of drugs administered to patients and drug manufacturers may wish to perform business analytics based on the purchase trends of the drugs. In the past, data privacy schemes [7], [10], [13], [17], [18], [19], [20], [25], [26], [27] have primarily focused on protecting individuals' information in sensitive datasets while allowing aggregate information on groups of individuals. Differential privacy [10] provides a model to quantify the disclosure risks by ensuring that the published statistical data does not depend on the presence or absence of a single individual's record in the dataset [10], [11].…”

“…Some data users (e.g., less privileged data analysts) may be allowed to obtain graph structure and aggregate information for a larger group size (e.g., the number of purchases of 'Psychiatric' drugs in the city of 'Pittsburgh') while some other more privileged data users may have access to the same information at smaller group sizes (e.g., the number of 'Psychiatric' drugs purchased in the zipcode '15206' within 'Pittsburgh'). While existing mechanisms [7], [10], [13], [19], [20], [26], [27], [31] have focused on protecting individual's sensitive values in datasets, this paper proposes a privacy-preserving data publishing mechanism addressing group privacy when aggregate information about groups of individuals can be sensitive and needs protection. Concretely, this paper makes the following key contributions: We first propose the notion of g -group differential privacy that provides guaranteed protection of aggregate information of a group of individuals in a given dataset.…”

Differentially Private Group-by Data Releasing Algorithm

“…Someone submits a query to a database, and it returns via anonymization algorithms an answer with noise [Mendonça et al 2017]. Occasionally, it is necessary to publish the data instead of statistics about it [Chen et al 2011].…”