Mobile social networks include a self-organizing system consisting of several mobile wireless nodes. Node behavior due to selfishness can significantly reduce performance on mobile social networks. A selfish node intends to use its resources only for itself and is reluctant to share them its neighbors. Therefore, identifying selfish nodes is very important to improve the performance of mobile social networks. In this paper, an improved method based on credit history for optimal detection of selfish nodes using the information distribution method is presented. The proposed method effectively recognizes selfish nodes and increases the cooperate motivation between network nodes. Among the various models, the fuzzy game theory model used to encourage the cooperation of selfish nodes is more efficient and about 98.5% of the packets have reached their destination on time, which in terms of energy consumption, delivery time and best use of the intermediate nodes outperforms other methods. If several intermediate nodes were willing to send packets, the node that had better conditions and provided better parameters and resources in return of the appropriate reward, was used based on the importance of the parameters in timely sending and lowest energy consumption. The remaining 1.5% of the nodes that were not suitable, removed from the network.