Participant Selection for Large- Scale Mobile Crowd Sensing System

Mobile crowd sensing (MCS) has become an emerging sensing paradigm for performing large-scale sensing tasks with the rapid increase in smartphones and big advancements in their embedded sensing technologies. One of the key challenges of large-scale MCS systems is effectively selecting appropriate participants from a huge user pool to perform various sensing tasks while satisfying certain constraints. This challenge becomes more complex when the capabilities of individual participants are diverse and unknown to the platform or participants are concerned about their privacy. In this talk, I will briefly discuss some recent results from my research group, aiming to address these challenges. First, to handle the multi-expertise of participants, we introduce a self-learning architecture that leverages the historical performing records of participants to learn different capabilities (both sensing probability and time delay) of participants. Formulating the participant selection problem as a combinational multi-armed bandit problem, we propose an online participant selection algorithm with both performance guarantee and bounded regret. Second, we investigate how to protect bid privacy in a temporally and spatially dynamic auction-based MCS system. We carefully design a scalable grouping-based privacy-preserving participant selection scheme, which places participants into small groups to satisfy their privacy requirements. It leverages Lagrange polynomial interpolation to perturb participants' bids for secure in-group bidding. Both theoretical analysis and real-life tracing data simulations verify the efficiency and security of the proposed solutions. Finally, I will conclude the talk with a lookout on future directions.

Earn Themes Credit in Energy To get Credit you must complete the assignment listed in Blackboard.