Preserving Distributed Spatio - Temporal Data Mining

Time-stamped location information is regarded as spatio-temporal data due to its time and space dimensions and, by its nature, is highly vulnerable to misuse. Privacy issues related to collection, use and distribution of individuals’ location information are the main obstacles impeding knowledge discovery in spatio-temporal data. Suppressing identifiers from the data does not suffice since movement trajectories can easily be linked to individuals using publicly available information such as home or work addresses. Yet another solution could be employing existing privacy preserving data mining techniques. However these techniques are not suitable since time-stamped location observations of an object are not plain, independent attributes of this object. Therefore, new privacy preserving data mining techniques are required to handle spatio-temporal data specifically. In this thesis, we propose a privacy preserving data mining technique and two preprocessing steps for data mining related to privacy preservation in spatio-temporal datasets: (1) Distributed clustering, (2) Centralized anonymization and (3) Distributed anonymization. We also provide security and efficiency analysis of our algorithms which shows that under reasonable conditions, achieving privacy preservation with minimal sensitive information leakage is possible for data mining purposes.