TY - JOUR

T1 - Confidentiality challenges in releasing longitudinally linked data

AU - Mitra, R.

AU - Blanchard, S.

AU - Dove, I.

AU - Tudor, C.

AU - Spicer, K.

PY - 2020

Y1 - 2020

N2 - Longitudinally linked household data allows researchers to analyse trends over time as well as on a cross-sectional level. Such analysis requires households to be linked across waves, but this increases the possibility of disclosure risks. We focus on an inter-wave disclosure risk specific to such data sets where intruders can make use of intimate knowledge gained about the household in one wave to learn new sensitive information about the household in future waves. We consider a specific way this risk could occur when households split in one wave, so an individual has left the household, and illustrate this risk using the Wealth and Assets survey. We also show that simply removing the links between waves may be insufficient to adequately protect confidentiality. To mitigate this risk we investigate two statistical disclosure control methods, perturbation and synthesis, that alter sensitive information on these households in the current wave. In this way no new sensitive information will be disclosed to these individuals, while utility should be largely preserved provided the SDC measures are applied appropriately. © 2020, University of Skovde. All rights reserved.

AB - Longitudinally linked household data allows researchers to analyse trends over time as well as on a cross-sectional level. Such analysis requires households to be linked across waves, but this increases the possibility of disclosure risks. We focus on an inter-wave disclosure risk specific to such data sets where intruders can make use of intimate knowledge gained about the household in one wave to learn new sensitive information about the household in future waves. We consider a specific way this risk could occur when households split in one wave, so an individual has left the household, and illustrate this risk using the Wealth and Assets survey. We also show that simply removing the links between waves may be insufficient to adequately protect confidentiality. To mitigate this risk we investigate two statistical disclosure control methods, perturbation and synthesis, that alter sensitive information on these households in the current wave. In this way no new sensitive information will be disclosed to these individuals, while utility should be largely preserved provided the SDC measures are applied appropriately. © 2020, University of Skovde. All rights reserved.

KW - Data confidentiality

KW - Disclosure risk

KW - Matching

KW - Perturbation

KW - Propensity score

KW - Synthetic data

KW - Linked data

KW - Perturbation techniques

KW - Current waves

KW - Household datum

KW - Sensitive informations

KW - Statistical disclosure Control

KW - Trends over time

KW - Risk assessment

M3 - Journal article

VL - 13

SP - 151

EP - 170

JO - Transactions on Data Privacy

JF - Transactions on Data Privacy

IS - 2

ER -