Credible nodes selection in mobile crowdsensing based on GAN

High-quality perception data is the basis of the operation of mobile crowdsensing platform. False data has a negative impact on the perception platform to draw wrong conclusions and cannot provide valuable help to service requesters. In order to solve this problem, we propose a trusted user selection framework based on generative adversarial network to select perceived users from the perspective of providing data, named PnGAN. The framework requires users to provide evaluation opinions on the current environment when uploading data, the evaluation opinion can not only reflect the current situation of the perceived environment, it can also be used as one of the data for user credibility evaluation. The analysis module in the framework processes the data uploaded by users, and apply the trust matrix to calculate the reputation value, determine user credibility based on reputation value. Experiments show that, compared with other algorithms for selecting trusted users, the average error rate of our method decreased by 30.62%, and the data quality improved by 15.32%.