Please use this identifier to cite or link to this item:
Title: Toolkit for privacy evaluation of geolocated data
Author: Gil García, Manel
Director: García Font, Víctor
Tutor: Salas Piñón, Julián
Keywords: privacy
Issue Date: Dec-2018
Publisher: Universitat Oberta de Catalunya (UOC)
Abstract: The rise of Location Based Systems during the last years has lead to a new problem, the data privacy of system's users. This Master's Final Project (TFM) aims to evaluate the privacy of the data obtained by sanitizing a dataset of spatiotemporal points through a series of sanitization methods. In order to evaluate the privacy of a geolocated dataset, we try to know how difficult it would be for an attacker to deduce sensitive information for a subject whose information is in the dataset, either by the deduction of points of interest (POIs) or by re-identifying a subject by linking traces. Perturbation, aggregation and swapping of traces have been evaluated as sanitization methods. In order to evaluate them, the sanitized datasets have been attacked by the following methods: home inferring, extraction of stays and begin-end location finder. The results must be considered from the point of view of the application that will be given to the sanitized data. If you want to keep all the aggregated data intact and use the sanitized dataset for the same purpose as the source one, swapping is the most powerful method, maintaining a good level of privacy. If the utility of the sanitized dataset does not take into account that the extracted data keeps aggregated data or that it can be used to, for example, create mobility maps -an unrealistic scenario-, perturbation is a very simple method to implement, although if different types of disturbance are not combined, an attacker could easily deduce the original locations.
Language: Spanish
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
mgilgarcia01TFM1218memoria.pdfMemoria del TFM17.04 MBAdobe PDFView/Open

This item is licensed under a Creative Commons License Creative Commons