Anomaly Detection Model Based on Extended Convolutional Neural Network
A DCNN-MiLSTM-based anomaly detection model is proposed to solve the problem that traditional network anomaly detection models are difficult to handle network traffic data with temporal characteristics.The timestamps of the original traffic data are redefined,and the overall features are extracted by using an expansive convolutional neural network,while the Mogrifier LSTM network is introduced for deeper mining of temporal information.Compared with other anomaly detection models,the DCNN-MiLSTM model achieves an accuracy of 99.12%,a recall of 98.94%,and a F1 of 99.03%,which are better than other common models in all metrics,and improves the ability of detecting anomalous network traffic data.The DCNN-MiLSTM model can better deal with traffic flows with temporal characteristics,capture the time in traffic data dependencies and trends in traffic data,and more effectively detect and identify anomalous data.