Privacy-Preserving Data Collection For Mobile Phone Sensing Tasks

Lack of reliable data is a major obstacle in some research works because users are unwilling to provide their own private data to any third parties directly. Since statistical inference is aimed to analyze the overall data of a well-defined group rather than a specific individual, the paradigm of privacy-preserving data collection scheme is proposed recently, which can motivate users to contribute their data to research works. In this paper, two probable properties that promote the success of sensing tasks are analyzed, and a fog-assisted data collection scheme for mobile phone sensing tasks is proposed. Sensitive measurements are particularly protected by obfuscating them with the group values, which not only provides anonymity for participants but also enables accurate data for the task provider. Especially, the dynamic change of participants is also considered. Theoretical analysis shows that this method achieves the desired security goals, and experiments are performed to demonstrate the efficiency and feasibility.