Network Traffic Intrusion Detection Based on Multi-Scale Convolution and Enhanced Temporal Convolution

In order to enhance the accuracy of network traffic intrusion detection and address the issues of complexity and lengthy training associated with traditional models for network traffic intrusion detection, this paper proposes a hybrid intrusion detection model that integrates Convolutional Neural Network (CNN) and Temporal Convolutional Network (TCN). The approach employs Multi-Scale One-Dimensional Convolution (MSC) and an improved Temporal Convolutional Network with Attention Mechanism (TCNAG) for spatial and temporal feature extraction learning. Finally, the model is trained and utilized for detection by combining it with a SoftMax classifier. TCNAG introduces an attention mechanism after the residual blocks of the traditional TCN and replaces the conventional fully connected layers with a global average pooling layer to avoid parameter redundancy, thus reducing the model's detection time. Experimental evaluations conducted on the UNSW-NB15 Dataset demonstrate that the proposed hybrid model exhibits lower time consumption and superior performance metrics compared to other commonly used algorithms.