[Objective] In order to explore the characteristics of rumor propagation, detect the source of rumor and reduce the harm caused by rumor.

[Methods] This paper adds the state of “rumor denier” to the traditional infectious disease model, sets the node state transformation rules according to the characteristics of rumor propagation in reality, and constructs a SIDR rumor propagation model considering the node interaction in social networks. Based on SIDR model, this paper proposes a rumor source detection algorithm and uses Beam search algorithm to optimize the detection model. On the basis of theoretical modeling, this paper selects typical real-world rumor cases to verify the model.

[Results] SIDR model can accurately simulate the propagation of rumors; rebutting humor from the source can inhibit the spread of rumor, and our source detection algorithm achieve high accuracy when applied in the early stage of rumor propagation.

[Limitations] This paper do not consider the dynamic changes of social networks in reality, and the representativeness of selected cases is limited. [Conclusions] The results provide guidance for the prediction of rumor propagation and the identification of rumor source.