Abstract: A network intrusion detection model based on multi-stage feature selection and CNN-GRU is proposed to address the problem of low accuracy of intrusion detection due to redundant features of network intrusion detection data.Firstly,for the feature redundancy of the data set,the PCC-RF feature selection algorithm is constructed by combining Pearson correlation coefficient and random forest for multistage feature selection and constructing the optimal feature subset.Then the CNN-GRU model is constructed by using the powerful extraction capability of convolutional neural network for spatial features and the excellent temporal feature extraction capability of gated recurrent units.Finally,the optimal feature subset is input into the CNN-GRU model for training.Experiments are conducted by using the UNSWNB15 dataset,and the experimental results show that the dataset,after the PCC-RF feature processing algorithm,has lower dimensionality and better results compared with other methods.The model detection accuracy reaches 84.72%.