微型电脑应用

为了提高网络入侵检测的效率，提出一种集成卷积神经网络（CNN）、双向门控循环单元（BiGRU）、注意力机制的改进卷积神经网络模型。所提出的网络入侵检测模型利用CNN提取网络流量数据的局部特征，通过BiGRU捕捉数据中的时序特征，并结合注意力机制对不同类型网络流量数据加权处理。将所提出的网络入侵检测模型应用于KDD Cup 99数据集，其对Normal、DOS、Probe、R2L、U2R5种攻击类型的检测准确率依次为92.8%、93.6%、92.7%、93.3%、91.4%。所提出的模型在DOS攻击类型下的4个评价指标均优于CNN模型和CNN-BiGRU模型，并通过消融实验验证了其相对于单向GRU,BiGRU在时序特征提取中更具优势。这为网络安全领域提供了一种可靠的网络入侵检测方法，具有一定的实用价值。

In order to improve the efficiency of network intrusion detection,an improved CNN model is proposed by integrating convolutional neural network(CNN),bidirectional gated recurrent units(BiGRU)and attention mechanism.The proposed model uses CNN to extract the local features of network traffic data,captures the time series features through BiGRU,and combines the attention mechanism to perform weighted processing on the different types of network traffic data.The proposed network intrusion detection model is applied to KDD Cup 99 dataset,and the detection accuracy of Normal,DOS,Probe,R2L and U2 Rattack types is 92.8%,93.6%,92.7%,93.3% and 91.4%,respectively.The proposed model is superior to the CNN model and the CNN-BiGRU model in four evaluation indexes under DOS attack types,and the ablation experiment verifies that BiGRU has more advantages in time series feature extraction than one-way GRU.This provides a reliable method for network intrusion detection in the field of network security,and has certain practical value.