JISE


  [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]


Journal of Information Science and Engineering, Vol. 39 No. 4, pp. 935-950


A Densely Stacked Attention Method for Cyberattack Detection


HAIXIA HOU1,*, DAOJUN LIANG2,3,*,
MINGQIANG ZHANG3,4 AND DONGFENG YUAN3,+
1College of Science and Information Science, Qingdao Agricultural University, 266109 China
2School of Information Science and Engineering, Shandong University, Qingdao, 266299 China
3Shandong Key Laboratory of Wireless Mobile Communication Technologies, Jinan, 250000 China
4School of Cyber Science and Engineering, Qufu Normal University, 273165 China
E-mail: hxhou@qau.edu.cn; {liangdaojun; mqzhang}@mail.sdu.edu.cn;

+dfyuan@sdu.edu.cn


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.


Keywords: cyberattack detection, intrusion detection system, deep learning, attention mechanism, data scarcity

  Retrieve PDF document (JISE_202304_14.pdf)