Carleton University - School of Computer Science Honours Project
Winter 2018
Creating Noise to Maintain Location Privacy: Truncation and Redistribution of Probability Mass
ABSTRACT
In privacy-preserving geographical information systems, there is a constant trade-off between utility and
privacy leakage of data. Adding noise drawn randomly from a probability density function to the location
data sacrifices some utility to maintain privacy. Truncating these probability density functions ensures
the noise added will always be within a predetermined range. This allows more control over the balance
between the utility of the service and the privacy to be maintained. This project implements various
mechanisms which truncate and redistribute this probability mass and studies their effect on the utility
of the service. Results of experiments run through the implementation of these mechanisms are
presented and discussed.