In this paper, we propose an improved CSR model for rumor spreading in mobile social networks. The dynamic equation of rumor spreading is modified to be suitable for user's habit in mobile social networks. In the acceptant probability model, negative and positive social reinforcements are considered. Furthermore, the people's accepting threshold for rumor accepting is taken into account. Analytically, a mean field theory is worked out by considering the influence of network topological structure as homogeneous. Under certain conditions, rumor spreads faster and wider in the new model than CSR rumor spreading model in homogeneous networks. Meanwhile, the multi-agent simulation results indicate that the information spreading process is sensitively dependent on initial conditions.