Network Intrusion Detection Using Temporal Convolutional Networks for Time-Series Traffic Modeling

With the continuous evolution of network attack methods, traditional intrusion detection systems have gradually exposed the problems of poor timeliness and insufficient modeling capabilities in identifying complex attack behaviors. In order to improve the system's ability to extract time series traffic features and the accuracy of attack type discrimination, this paper proposes a network intrusion detection method based on Temporal Convolutional Network (TCN). The constructed model combines dilated convolution and causal convolution mechanisms to achieve large receptive field coverage while maintaining sequence order, which is suitable for modeling tasks of complex time series data. The model was experimentally verified on standard datasets NSL-KDD and CICIDS-2017, and its performance was compared with mainstream methods such as CNN, LSTM and Transformer. The results show that the proposed method achieves optimal performance in multiple key indicators such as accuracy, recall rate and F1 value, and has good detection stability and anti-interference ability. The study shows that TCN has broad application prospects in the field of network security time series modeling, and provides a feasible path for building efficient and deployable intelligent intrusion detection systems.