Anonymizing mobility data using semantic cloaking

Abstract

The prevalence of mobile phones has led to an explosion in the amounts of human mobility data stored in the cloud. It has been shown that seemingly anonymized location datasets are highly susceptible for re-identification, and may not provide enough privacy protection. In this paper we quantitatively show how semantic cloaking, the application of semantic labeling to achieve anonymization, can improve the privacy of a mobility dataset for use cases where location coordinates can be replaced by semantic categories. We develop a semantic labeling framework, apply it and evaluate it using the dataset uniqueness (ε) measure. Our experiments show an improvement in uniqueness ranging between two- and twenty two-fold in comparison to the original, naively anonymized, dataset.