Trust-aware truth discovery with Long-term Vehicle Reputation for Internet of Vehicles Crowdsensing

In Internet of Vehicles (IoV), vehicle-based crowdsensing provides various significant data, such as weather condition and GPS data, etc. However, due to different source quality, the sensed data of vehicles may vary from the ground truth. Truth discovery is usually used to analyze conflicting data, and it traditionally estimates source quality only from the current task. Aiming to take the long-term reputation into consideration, in this paper, we propose a trust-aware model. Specifically, we first propose a new model to define vehicle reliability. Instead of using weight as reliability, we define source reliability by its contribution to the loss function. Then, we propose a novel trust-aware model to accumulate long-term reputation of vehicle from its history tasks. Finally, based on vehicle reputation, initial weights are assigned for the future task. Experiments conducted on weather conditions and GPS datasets demonstrate that our trust-aware model can save execution time and achieve higher accuracy.