Differentially Private and Skew-Aware Spatial Decompositions for Mobile Crowdsensing

Jong Seon Kim, Yon Dohn Chung, Jong Wook Kim

Research output: Contribution to journalArticle

Abstract

Mobile Crowdsensing (MCS) is a paradigm for collecting large-scale sensor data by leveraging mobile devices equipped with small and low-powered sensors. MCS has recently received considerable attention from diverse fields, because it can reduce the cost incurred in the process of collecting a large amount of sensor data. However, in the task assignment process in MCS, to allocate the requested tasks efficiently, the workers need to send their specific location to the requester, which can raise serious location privacy issues. In this paper, we focus on the methods for publishing differentially a private spatial histogram to guarantee the location privacy of the workers. The private spatial histogram is a sanitized spatial index where each node represents the sub-regions and contains the noisy counts of the objects in each sub-region. With the sanitized spatial histograms, it is possible to estimate approximately the number of workers in the arbitrary area, while preserving their location privacy. However, the existing methods have given little concern to the domain size of the input dataset, leading to the low estimation accuracy. This paper proposes a partitioning technique SAGA (Skew-Aware Grid pArtitioning) based on the hotspots, which is more appropriate to adjust the domain size of the dataset. Further, to optimize the overall errors, we lay a uniform grid in each hotspot. Experimental results on four real-world datasets show that our method provides an enhanced query accuracy compared to the existing methods.

Original languageEnglish
JournalSensors (Basel, Switzerland)
Volume18
Issue number11
DOIs
Publication statusPublished - 2018 Oct 30

Fingerprint

privacy
Privacy
histograms
Decomposition
decomposition
sensors
Sensors
grids
Mobile devices
preserving
costs
Costs and Cost Analysis
Equipment and Supplies
estimates
Datasets
Costs

Keywords

  • differential privacy
  • histograms
  • mobile crowdsensing
  • spatial databases

ASJC Scopus subject areas

  • Analytical Chemistry
  • Atomic and Molecular Physics, and Optics
  • Biochemistry
  • Instrumentation
  • Electrical and Electronic Engineering

Cite this

Differentially Private and Skew-Aware Spatial Decompositions for Mobile Crowdsensing. / Kim, Jong Seon; Chung, Yon Dohn; Kim, Jong Wook.

In: Sensors (Basel, Switzerland), Vol. 18, No. 11, 30.10.2018.

Research output: Contribution to journalArticle

@article{a4e0a9adaa0d4fe591c5484f83896d11,
title = "Differentially Private and Skew-Aware Spatial Decompositions for Mobile Crowdsensing",
abstract = "Mobile Crowdsensing (MCS) is a paradigm for collecting large-scale sensor data by leveraging mobile devices equipped with small and low-powered sensors. MCS has recently received considerable attention from diverse fields, because it can reduce the cost incurred in the process of collecting a large amount of sensor data. However, in the task assignment process in MCS, to allocate the requested tasks efficiently, the workers need to send their specific location to the requester, which can raise serious location privacy issues. In this paper, we focus on the methods for publishing differentially a private spatial histogram to guarantee the location privacy of the workers. The private spatial histogram is a sanitized spatial index where each node represents the sub-regions and contains the noisy counts of the objects in each sub-region. With the sanitized spatial histograms, it is possible to estimate approximately the number of workers in the arbitrary area, while preserving their location privacy. However, the existing methods have given little concern to the domain size of the input dataset, leading to the low estimation accuracy. This paper proposes a partitioning technique SAGA (Skew-Aware Grid pArtitioning) based on the hotspots, which is more appropriate to adjust the domain size of the dataset. Further, to optimize the overall errors, we lay a uniform grid in each hotspot. Experimental results on four real-world datasets show that our method provides an enhanced query accuracy compared to the existing methods.",
keywords = "differential privacy, histograms, mobile crowdsensing, spatial databases",
author = "Kim, {Jong Seon} and Chung, {Yon Dohn} and Kim, {Jong Wook}",
year = "2018",
month = "10",
day = "30",
doi = "10.3390/s18113696",
language = "English",
volume = "18",
journal = "Sensors (Switzerland)",
issn = "1424-8220",
publisher = "Multidisciplinary Digital Publishing Institute (MDPI)",
number = "11",

}

TY - JOUR

T1 - Differentially Private and Skew-Aware Spatial Decompositions for Mobile Crowdsensing

AU - Kim, Jong Seon

AU - Chung, Yon Dohn

AU - Kim, Jong Wook

PY - 2018/10/30

Y1 - 2018/10/30

N2 - Mobile Crowdsensing (MCS) is a paradigm for collecting large-scale sensor data by leveraging mobile devices equipped with small and low-powered sensors. MCS has recently received considerable attention from diverse fields, because it can reduce the cost incurred in the process of collecting a large amount of sensor data. However, in the task assignment process in MCS, to allocate the requested tasks efficiently, the workers need to send their specific location to the requester, which can raise serious location privacy issues. In this paper, we focus on the methods for publishing differentially a private spatial histogram to guarantee the location privacy of the workers. The private spatial histogram is a sanitized spatial index where each node represents the sub-regions and contains the noisy counts of the objects in each sub-region. With the sanitized spatial histograms, it is possible to estimate approximately the number of workers in the arbitrary area, while preserving their location privacy. However, the existing methods have given little concern to the domain size of the input dataset, leading to the low estimation accuracy. This paper proposes a partitioning technique SAGA (Skew-Aware Grid pArtitioning) based on the hotspots, which is more appropriate to adjust the domain size of the dataset. Further, to optimize the overall errors, we lay a uniform grid in each hotspot. Experimental results on four real-world datasets show that our method provides an enhanced query accuracy compared to the existing methods.

AB - Mobile Crowdsensing (MCS) is a paradigm for collecting large-scale sensor data by leveraging mobile devices equipped with small and low-powered sensors. MCS has recently received considerable attention from diverse fields, because it can reduce the cost incurred in the process of collecting a large amount of sensor data. However, in the task assignment process in MCS, to allocate the requested tasks efficiently, the workers need to send their specific location to the requester, which can raise serious location privacy issues. In this paper, we focus on the methods for publishing differentially a private spatial histogram to guarantee the location privacy of the workers. The private spatial histogram is a sanitized spatial index where each node represents the sub-regions and contains the noisy counts of the objects in each sub-region. With the sanitized spatial histograms, it is possible to estimate approximately the number of workers in the arbitrary area, while preserving their location privacy. However, the existing methods have given little concern to the domain size of the input dataset, leading to the low estimation accuracy. This paper proposes a partitioning technique SAGA (Skew-Aware Grid pArtitioning) based on the hotspots, which is more appropriate to adjust the domain size of the dataset. Further, to optimize the overall errors, we lay a uniform grid in each hotspot. Experimental results on four real-world datasets show that our method provides an enhanced query accuracy compared to the existing methods.

KW - differential privacy

KW - histograms

KW - mobile crowdsensing

KW - spatial databases

UR - http://www.scopus.com/inward/record.url?scp=85055836378&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85055836378&partnerID=8YFLogxK

U2 - 10.3390/s18113696

DO - 10.3390/s18113696

M3 - Article

C2 - 30380798

AN - SCOPUS:85055836378

VL - 18

JO - Sensors (Switzerland)

JF - Sensors (Switzerland)

SN - 1424-8220

IS - 11

ER -