In the current information age, image tampering detection technology is crucial to ensure the integrity and authenticity of digital media, and remote image tampering detection technology combined with deep neural networks has become a research hotspot. This paper adopts convolutional neural network as the main detection tool, and on the improved DPN network model, the feature fusion module based on the attention mechanism is used to fuse the two features in this paper. In this way, the image tampering detection technique based on dual-stream feature fusion is proposed in this paper. The precision, recall and F value of the detection algorithm in this paper are better than the comparison algorithm. When the image compression quality factor is reduced to 20, the precision rate, recall rate, and F value of this paper’s algorithm do not appear to be greatly reduced, and the reduction is only 0.028, 0.041, and 0.042. This paper’s image tampering detection algorithm, which fuses the frequency domain branching module and the attention mechanism feature fusion module, has a higher detection efficiency. And the Accuracy rate, Recall rate and F Value of this paper’s algorithm on image level detection are 17.8%, 15.3% and 16.3% higher than that of DCT algorithm respectively. In conclusion, the remote image tampering technique combined with deep neural network provides an effective solution to ensure the authenticity and integrity of images.