K-Skyband query answering with differential privacy1

Ling Chen, Ting Yu, Rada Chirkova

Research output: Contribution to journalArticle

Abstract

Given a set of multi-dimensional points, a k-skyband query retrieves those points dominated by no more than k other points. k-skyband queries are an important type of multi-criteria analysis with diverse applications in practice. In this paper, we investigate techniques to answer k-skyband queries with differential privacy.We first propose a general technique BBS-Priv, which accepts any differentially private spatial decomposition tree as input and leverages data synthesis to answer k-skyband queries privately. We then show that, though quite a few private spatial decomposition trees are proposed in the literature, they are mainly designed to answer spatial range queries. Directly integrating them with BBS-Priv would introduce too much noise to generate useful k-skyband results. To address this problem, we propose a novel spatial decomposition technique k-skyband tree specially optimized for k-skyband queries, which partitions data adaptively based on the parameter k and performs finer partitions on the regions that are likely to contain k-skyband results.We further propose techniques to generate a k-skyband tree over spatial data that satisfies differential privacy, and combine BBS-Priv with the private k-skyband tree to answer k-skyband queries.We conduct extensive experiments based on two real-world datasets and three synthetic datasets that are commonly used for evaluating k-skyband queries. The results show that the proposed scheme significantly outperforms existing differentially private spatial decomposition schemes and achieves high utility when privacy budgets are properly allocated.

Original languageEnglish
Pages (from-to)647-676
Number of pages30
JournalJournal of Computer Security
Volume26
Issue number5
DOIs
Publication statusPublished - 1 Jan 2018

Fingerprint

Bulletin boards
Decomposition
Experiments

Keywords

  • Adaptive spatial decomposition
  • Differential privacy
  • K-Skyband query

ASJC Scopus subject areas

  • Software
  • Safety, Risk, Reliability and Quality
  • Hardware and Architecture
  • Computer Networks and Communications

Cite this

K-Skyband query answering with differential privacy1. / Chen, Ling; Yu, Ting; Chirkova, Rada.

In: Journal of Computer Security, Vol. 26, No. 5, 01.01.2018, p. 647-676.

Research output: Contribution to journalArticle

Chen, Ling ; Yu, Ting ; Chirkova, Rada. / K-Skyband query answering with differential privacy1. In: Journal of Computer Security. 2018 ; Vol. 26, No. 5. pp. 647-676.
@article{a0b3c37834414128bf2e391a7b24a351,
title = "K-Skyband query answering with differential privacy1",
abstract = "Given a set of multi-dimensional points, a k-skyband query retrieves those points dominated by no more than k other points. k-skyband queries are an important type of multi-criteria analysis with diverse applications in practice. In this paper, we investigate techniques to answer k-skyband queries with differential privacy.We first propose a general technique BBS-Priv, which accepts any differentially private spatial decomposition tree as input and leverages data synthesis to answer k-skyband queries privately. We then show that, though quite a few private spatial decomposition trees are proposed in the literature, they are mainly designed to answer spatial range queries. Directly integrating them with BBS-Priv would introduce too much noise to generate useful k-skyband results. To address this problem, we propose a novel spatial decomposition technique k-skyband tree specially optimized for k-skyband queries, which partitions data adaptively based on the parameter k and performs finer partitions on the regions that are likely to contain k-skyband results.We further propose techniques to generate a k-skyband tree over spatial data that satisfies differential privacy, and combine BBS-Priv with the private k-skyband tree to answer k-skyband queries.We conduct extensive experiments based on two real-world datasets and three synthetic datasets that are commonly used for evaluating k-skyband queries. The results show that the proposed scheme significantly outperforms existing differentially private spatial decomposition schemes and achieves high utility when privacy budgets are properly allocated.",
keywords = "Adaptive spatial decomposition, Differential privacy, K-Skyband query",
author = "Ling Chen and Ting Yu and Rada Chirkova",
year = "2018",
month = "1",
day = "1",
doi = "10.3233/JCS-171101",
language = "English",
volume = "26",
pages = "647--676",
journal = "Journal of Computer Security",
issn = "0926-227X",
publisher = "IOS Press",
number = "5",

}

TY - JOUR

T1 - K-Skyband query answering with differential privacy1

AU - Chen, Ling

AU - Yu, Ting

AU - Chirkova, Rada

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Given a set of multi-dimensional points, a k-skyband query retrieves those points dominated by no more than k other points. k-skyband queries are an important type of multi-criteria analysis with diverse applications in practice. In this paper, we investigate techniques to answer k-skyband queries with differential privacy.We first propose a general technique BBS-Priv, which accepts any differentially private spatial decomposition tree as input and leverages data synthesis to answer k-skyband queries privately. We then show that, though quite a few private spatial decomposition trees are proposed in the literature, they are mainly designed to answer spatial range queries. Directly integrating them with BBS-Priv would introduce too much noise to generate useful k-skyband results. To address this problem, we propose a novel spatial decomposition technique k-skyband tree specially optimized for k-skyband queries, which partitions data adaptively based on the parameter k and performs finer partitions on the regions that are likely to contain k-skyband results.We further propose techniques to generate a k-skyband tree over spatial data that satisfies differential privacy, and combine BBS-Priv with the private k-skyband tree to answer k-skyband queries.We conduct extensive experiments based on two real-world datasets and three synthetic datasets that are commonly used for evaluating k-skyband queries. The results show that the proposed scheme significantly outperforms existing differentially private spatial decomposition schemes and achieves high utility when privacy budgets are properly allocated.

AB - Given a set of multi-dimensional points, a k-skyband query retrieves those points dominated by no more than k other points. k-skyband queries are an important type of multi-criteria analysis with diverse applications in practice. In this paper, we investigate techniques to answer k-skyband queries with differential privacy.We first propose a general technique BBS-Priv, which accepts any differentially private spatial decomposition tree as input and leverages data synthesis to answer k-skyband queries privately. We then show that, though quite a few private spatial decomposition trees are proposed in the literature, they are mainly designed to answer spatial range queries. Directly integrating them with BBS-Priv would introduce too much noise to generate useful k-skyband results. To address this problem, we propose a novel spatial decomposition technique k-skyband tree specially optimized for k-skyband queries, which partitions data adaptively based on the parameter k and performs finer partitions on the regions that are likely to contain k-skyband results.We further propose techniques to generate a k-skyband tree over spatial data that satisfies differential privacy, and combine BBS-Priv with the private k-skyband tree to answer k-skyband queries.We conduct extensive experiments based on two real-world datasets and three synthetic datasets that are commonly used for evaluating k-skyband queries. The results show that the proposed scheme significantly outperforms existing differentially private spatial decomposition schemes and achieves high utility when privacy budgets are properly allocated.

KW - Adaptive spatial decomposition

KW - Differential privacy

KW - K-Skyband query

UR - http://www.scopus.com/inward/record.url?scp=85051964886&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85051964886&partnerID=8YFLogxK

U2 - 10.3233/JCS-171101

DO - 10.3233/JCS-171101

M3 - Article

VL - 26

SP - 647

EP - 676

JO - Journal of Computer Security

JF - Journal of Computer Security

SN - 0926-227X

IS - 5

ER -