Information transfer is the basic feature of the network. Accordingly, an adaptive community detection method based on information transfer and density peaks is proposed. Firstly, the trust degree function between nodes and neighbors is defined, and each node independently spreads amount of information to the network based on the trust degree. After the diffusion, the total information amount of the node is the density of the density peaks. The distance between the nodes in the network is replaced by the reciprocal of the information amount of the destination node. Then, a method is proposed that can automatically select core nodes which are divided into different communities, and the remaining nodes are allocated to the community of the closest core node. The algorithm has the advantage that no additional parameters are needed and the internal structure of the community can be found. The experimental results show that.