摘要
针对传统单一深度学习模型在网络入侵检测中的局限性,提出一种基于CNN和LSTM特征融合的网络入侵检测方法.该方法通过CNN提取网络流量的空间特征,利用LSTM捕获时序依赖关系,将二者特征融合后引入自注意力机制以强化关键特征权重,最后通过Softmax层实现多类别入侵行为的精准分类.在NSL-KDD数据集的实验中,模型的F1-score达到98.3%,较传统单一深度学习模型在检测精度上有显著提升,验证了特征融合与注意力机制结合的有效性.
Aiming at the limitations of traditional single deep learning models in network intrusion detection,this paper proposes a network intrusion detection method based on the feature fusion of CNN and LSTM.Specifically,the method uses CNN to extract the spatial features of network traffic and leverages LSTM to capture temporal dependencies.After fusing these two types of features,a self attention mechanism is introduced to enhance the weights of key features.Finally,a Softmax layer is employed to achieve accurate classification of multi-category intrusion behaviors.In the experiments on the NSL-KDD dataset,the model achieves an F1-score of 98.3%,which shows a significant improvement in detection accuracy compared with traditional single deep learning models.This verifies the effectiveness of combining feature fusion with the attention mechanism.
作者
沈锐
SHEN Rui(Sichuan Tourism University,Chengdu Sichuan 610100,China)
出处
《太原师范学院学报(自然科学版)》
2025年第4期45-49,共5页
Journal of Taiyuan Normal University(Natural Science Edition)
关键词
网络入侵检测
特征融合
自注意力机制
network intrusion detection
feature fusion
self-attention mechanism
