Abstract: Studies on Mobile Crowdsensing(MCS) are generally based on the premise that the Edge Server(ES) or Cloud Server(CS) is trusted, which can not effectively protect the privacy of participants while improving the perceived data quality. A Privacy-Enhanced Multi-Task Assignment(PEMTA) mechanism based on a semi-trusted execution environment is proposed. The tasks are clustered based on Hilbert curve characteristics. The adjacent ESs are combined with the homomorphic characteristics of the Paillier encryption system for cooperation. The best set of participants is selected for each task based on the matching degree of participants and tasks to complete the perception task without disclosing the privacy of participants. The greedy conflict elimination algorithm is designed, and the conflict tasks are graded according to the task commission. The best replacement participants are selected for the conflict tasks based on the divided task level to resolve the matching participant conflict caused by multi-task allocation. The dynamic reputation value update algorithm is used to dynamically update the reputation value of participants by quantifying the deviation between the perception data submitted by participants and the aggregated data, alleviating data quality loss caused by malicious attacks. The experimental results show that the PEMTA mechanism performs satisfactorily in anti-malicious attacks. The perceived data quality and task completion rate increase by 18.14% and 15.47%, respectively, compared with similar multi-task allocation mechanisms on average.