Using network analysis for improving privacy in crowd sensing systems

Crowd sensing (CS) is a new paradigm that takes advantage of the massive use of smart phones, and the increasing computational capabilities of these devices for collecting and reporting sensed data. However, different issues must be addressed in order to guarantee the success of CS systems in real scenarios. In this paper, we explore the issue associated to the disclosing of participants´ location in CS systems, and how this concern may affect the users´ participation, jeopardizing the system. In order to address this concern, we propose the use of network analysis techniques to increase the participants´ privacy protection. The proposed mechanism uses the Kernighan-Lin and spectral-partitioning algorithms for improving the privacy protection provided by the point-of-interest mechanism. In addition, this paper presents a set of experiments that measures the effectiveness of the proposed mechanism showing that by using the spectral partitioning algorithm the anonymization degree is increased in a 86%.