Users have different privacy information disclosure requirements when they submit private data to service composition, and the composition should support the verification of users' privacy requirements. This paper puts forward a flexible method for users to produce privacy requirement specifications. Users can define the sensitivity of private data and its usage in different situations, and restrict the member services that can use private data with sensitivity-reputation function. The simplification and universality of the privacy requirements can be improved by using this method. The process first establishes privacy data item relations by using the privacy data item dependency graph (PDIDG), then models the service composition with privacy open workflow net (POWFN), and at last, makes sure whether service composition meets the user's privacy requirements by privacy requirements verification algorithm. An example is provided to illustrate the effectiveness of the method, and experiment analysis on the performance of the verification algorithm is carried out at the end of paper.