JISE

Cyberattack Detection plays a vital role in network security and is an important means to maintain network security. In order to enhance the security and improve the detection ability of malicious intrusion behavior in the network, this paper proposes a multi-layer Dense Attention (Denseformer) model. The model is composed of multiple transformerlike structures, and each layer is stacked of multiple encoder and decoder sub-layers. The encoder and decoder include self-attention and cross-attention mechanisms, and their features are obtained by cross-fusion of multi-branch structures. By sharing information among multiple encoder-decoder layers, Denseformer can use the attention mechanism to process unserialized input source samples. On the whole, Denseformer is like an attention network embedded on the dense layer, making it easier to handle correlations between features. By stacking the encoding and decoding modules with attention, Denseformer has better generalization performance than other models, thereby improving its cyberattack detection accuracy. The experimental results show that, without other complex training techniques, the proposed method achieves 85.65% on the NSL-KDD dataset.