For millions of people are joining mobile social networks every day, it is critical to establish trust among the indirectly connected users for improving the quality of mobile social network services and enforcing the security for them. A dynamic trust evaluation model was proposed to calculate and update trust metrics between one user and another user, which was based on direct trust, recommendation trust, trust risk function, motivate mechanism and the framework of mobile social network. Experiments with a dataset from a real social network Epinions.com show that the predicting trust metrics is accurate and the model effectively resist the attacks of malicious nodes.