Abstract

Location-aware technologies and smart phones are fast growing in usage and adoption as a medium of service request and delivery of daily activities. However, the increasing usage of these technologies has birthed new challenges that needs to be addressed. Privacy protection and the need for disaggregate mobility data for transportation modelling needs to be balanced for applications and academic research. This dissertation focuses on developing modern privacy mechanisms that seek to satisfy requirements on privacy and data utility for fine-grained travel behavioural modelling applications using large-scale mobility data. To accomplish this, we review the challenges and opportunities that are needed to be solved in order to harness the full potential of “Big Transportation Data”. Also, we perform a quantitative evaluation on the degree of privacy that are provided by popular location anonymization techniques when undertaken on sensitive location data (i.e. homes, offices) of a travel survey. As a step to solve the trade-off between privacy and utility, we develop a differentially-private generative model for simultaneously synthesizing both socio-economic attributes and sequences of activity diary. Adversarial attack models are proposed and tested to evaluate the effectiveness of the proposed system against privacy attacks. The results show that datasets from the developed privacy enhancing system can be used for travel behavioural modelling with satisfactory results while ensuring an acceptable level of privacy.