Abstract: With the development of Internet,the network environment is becoming more complex,and the resulting network security problems continue to emerge,so the protection of network security becomes an important research topic.Aiming at the problems of unbalanced traffic data collected in real network environment and inaccurate feature representation extracted by traditional machine learning methods,this paper proposes an intrusion detection method based on Borderline-SMOTE and dual attention.Firstly,this method performs Borderline-SMOTE oversampling on the intrusion data to solve the problem of data imbalance,and uses the advantages of convolutional networks for image feature extraction to convert 1D flow data into grayscale images.Then it updates the low-dimensional features from the channel dimension and the spatial dimension to obtain a more accurate feature representation respectively.Finally,it uses the Softmax classifier to classify and predict traffic data.The simulation experiments of the proposed method have been verified on the NSL-KDD data set,and the accuracy reaches 99.24%.Compared with other commonly used methods,it has a higher accuracy.

Key words: Borderline-SMOTE, Double Attention, Intrusion detection, Network security, Unbalanced problems