Intrusion Detection Method Based on RB-SMOTE and Double Attention

Aiming at the problems of unbalanced traffic data collected in the real network environment and inaccurate feature representation of traditional machine learning methods, this paper proposes an intrusion detection method based on RBSMOTE and double Attention. Firstly, based on a refined Borderline-SMOTE method (RB-SMOTE), the intrusion data is oversampled to solve the problem of data imbalance, and the advantages of the convolutional network in image feature extraction are used to extract Convert one-dimensional flow data into grayscale images. Then, the low-dimensional features are dimensionally updated from the channel dimension and the spatial dimension through the dual-attention network to obtain a more accurate feature representation. Finally, the softmax classifier is used to classify and predict the traffic data. This method has been verified on the NSL-KDD dataset, and its accuracy rate reaches 99.24%, which is higher than other commonly used methods.