The advantages of cloud computing such as open resource access interface and centralized management of massive resources invariably increase the possibility of cloud resources and other cloud service users(CSU) being threatened by untrustworthy CSUs. To effectively reduce the occurrence of such situations that threats to cloud security, it is necessary to implement separate controls for CSUs with different trust levels, and the discussion of the CSU behavioural trust sorting method is crucial. Therefore, a hesitant fuzzy CSU behaviour trust sorting approach with the function of rewarding good and penalizing bad is proposed. Based on the TOPSIS-Sort-C framework, the method uses a hesitant fuzzy set(HFS) to characterize the different CSU behavioural trust data from different sources, takes quantiles of each CSU behavioural trust indicator as the sorting thresholds. By non-linearly deflating the CSU behavioural data at different trust levels, the majorant degree of trust or distrust(MDT, MDD) of CSU behavioural reflecting reward or penalty are obtained to make the CSU behavioural trust level more intuitive while further expanding the graded differences. Finally, we confirm the effectiveness of the proposed method and the advantages of majorant reward or penalty by a comparison analysis with an example of CSU behaviour trust sorting in an internet company.