Reliability-driven Task Assignment in Vehicular Crowdsourcing: A Matching Game

Vehicular crowdsourcing is an emerging mobile sensing-based paradigm in which a service platform manages the allocation of data collection and processing tasks to vehicles according to their itinerary, expected response time, and optimal reward. However, selfish or malicious vehicles may take advantage of the platform to maximize their profit or even attack the deployed sensing applications and compromise their decisions by transmitting unreliable data, which can degrade the quality of delivered services and negatively affect the platform reward and reputation. In this paper, we design a task assignment mechanism for vehicular crowdsourcing based on vehicles' reliability to protect the platform from such threats. First, a beta reputation system is adopted for probabilistic reliability assessment according to vehicles' behavior. Then, a new multi-dimensional task assignment problem, which proves to be NP-complete, is modeled as a many-to-one matching game between the platform and vehicles by defining reliability and reward-based preference functions. Finally, a distributed matching algorithm that aims at maximizing the platform reward by assigning the tasks to reliable participants is presented. The mechanism also achieves privacy-preservability by enabling vehicles to engage in the task allocation process without necessarily unveiling their sensitive information. Results show that the proposed mechanism increases the quality of crowdsourcing services by assigning the tasks to reliable and benign vehicles, and adheres to the scalability requirements of the system.