Location-history partitioning algorithms for privacy in non-trusted geo-social networks

Abstract

Due to the rapid development of mobile device technology in the past couple of decades, mobile devices are playing a more and more important part in our daily life. Many mobile services along with mobile devices have integrated into our activities or even reshaped our lifestyle. Location services are one of the main mobile services being widely used. One can share ones location to get to know information nearby, and it could be shared with friends in social media. New mobile applications are showing up at an amazing speed, together with that, the usage of location data is a privacy threat. If too much information is shared, the user's movements could be predicted; highly privacy sensitive locations, such as home location of user, could be leaked. Many location based applications, such as geo-social networks (GSN), use Location Servers to store user position information. However, since GSN providers may not be fully trustworthy or may not be able to protect user data, users may not want to store all of their privacy-sensitive location information with a single provider. Therefore, this thesis focuses on developing and evaluating methods to partition location data among multiple servers as similarly attempted in other approaches. In this thesis, we try to partition location data to achieve privacy protection. We have studied a range of mobility modeling methods that consider the different fundamental dimensions of the location data, i.e., spatial, temporal, and semantic, as well as their combinations. Inspired by those methods, we have proposed partitioning methods to increase privacy protection. Furthermore, a couple of other partition methods, which are combinations of spatial, temporal and semantic, are implemented. Eventually all the partition methods are evaluated with our data.