【摘要】 在数字化时代，网络安全的重要性愈发显著，尤其是网络入侵检测（Network Intrusion Detection, NID）作为保障网络安全的核心环节。传统NID系统在检测效率与准确性方面表现不足，难以适应日益复杂多变的网络环境，因此需要创新解决方案以强化网络安全防护。文章提出了SE-TransNet：一种结合SENet的新型Transformer网格入侵检测方法。引入Transformer架构，增强模型处理长序列数据的能力，提升对网络流量时序特征的捕捉；将选择性注意力机制（SENet）集成于Transformer，通过自适应调整通道权重，强化特征表示，突出关键特征，加快响应速度。在CIC-IDS2017数据集上的实验显示，SE-TransNet实现了99.37%的准确率，较Transformer-CNN等模型提高0.50%～9.13%，证明了其在提升NID效率和准确性方面的优势。研究结果为网络安全提供了一定的技术支撑。更多还原

【Abstract】 In the digital age, the importance of cybersecurity has become increasingly significant, especially network intrusion detection(NID) which serves as the core component in ensuring network security. Traditional NID systems have shortcomings in detection efficiency and accuracy, making them difficult to adapt to the increasingly complex and dynamic network environment. Therefore, innovative solutions are urgently needed to strengthen network security protection. This paper proposes, SE-TransNet a novel Transformer-based Network Intrusion Detection method integrating SENet. By introducing the Transformer structure, the model’s ability to handle long sequence data is improved, and the capture of network traffic temporal features is improved. The integration of SENet into the Transformer allows for adaptive adjustment of channel weights, reinforcing feature representation, highlighting key features, and accelerating response speed. Experimental results on the CIC-IDS2017 dataset demonstrate that SETransNet achieved an accuracy rate of 99.37%, which is 0.50%～9.13% higher than models such as TransformerCNN, demonstrating its advantages in improving the efficiency and accuracy of NID. The research results provide certain technical support for cybersecurity.更多还原