Mobile social sensing leverages a large group of individuals having mobile devices capable of sensing and computing to perform intelligence sensing tasks. To perform smart participant selection (i.e. task assignment) for mobile social sensing, most of the cloud-based platforms often require mobile users to report their personal information, such as sensing cost, location and sensing quality. Therefore, the users might suffer from potential privacy breaches during the participant selection phase especially when they are not selected for performing the tasks. Existing solutions based on participant grouping can resolve this privacy leakage by leveraging secure sharing or group bidding within groups. However, the group formation problem has not be well studied, and the communication overhead over formed groups is usually ignored. To address these issues, in this paper, we consider a new set of privacy-preserving grouping problems over edge clouds to minimize the communication cost at edge clouds during secure sharing/bidding, while satisfying each participant's privacy requirement. Various grouping schemes are carefully designed to fulfill the optimization goal for two different scenarios: tree-based hierarchical edge clouds and graph-based interconnected edge clouds. Extensive simulations over both synthetic and real-life datasets are conducted to confirm the efficiency of all proposed schemes.