Companies providing Location Based Services collect large amounts of data that give them information about user movements. The collected data is then used for creating trajectories that can be used for activity mining which is then used to improve user experiences, such as personalised recommendations. Often these accumulated user movement trajectories are published by companies under pressure or shared with third parties for business. Such trajectories can contain users' sensitive visit information which can be used to extract behavioural patterns. Sharing or releasing information on such sensitive location visits makes them vulnerable to inference attacks. We have designed a methodology that includes both the generalization and suppressions to anonymise trajectories in order to protect from inference attacks based on association rules derived from data. Experimental results conducted on the real datasets show that the proposed algorithms can efficiently prevent inference attacks.
In general, we study, the problem of privacy leakage caused by the accumulation and sharing of numerous trajectory information of moving objects. Based on the methodology of anonymising trajectories. a dummy trajectory-based trajectory algorithm is proposed. In this algorithm, a heuristic rule is formulated that takes the comprehensive measure of trajectory similarity and location diversity dummy trajectories are selected. The dummy trajectories are then used to hide real trajectories and sensitive information. A number of different techniques to optimise the execution of anonymising algorithms have been studied. One such technique, based on a directed graph and grid-based map will be discussed. An empirical study on using real data sets is in progress.
I am a senior lecturer at school of IT, WelTec. My research interests are, concurrent data structures, cryptographic authentication and privacy. I teach courses in software engineering and cyber security.