Intrusion Detection Model Using Temporal Convolutional Network Blend Into Attention Mechanism

In order to improve the ability detect network attacks, traditional intrusion detection models often used convolutional neural networks encode spatial information or recurrent obtain temporal features of data. Some combined two methods extract spatio-temporal features. However, these approaches separate and learned insufficiently. This paper presented an improved model based on (TCN) attention mechanism. The causal dilation convolution can capture dependencies residual blocks allow transfer in a cross-layered manner, enabling in-depth learning. Meanwhile, mechanism enhance model's relevant anomalous different attacks. Finally, this compared results KDD CUP99 UNSW-NB15 datasets. Besides, authors apply video surveillance attack scenarios. result shows that has advantages evaluation metrics.