微型电脑应用

为了提高大数据环境下网络入侵检测精度，提出一种基于改进GAN(generative adversarial network)与改进Bi-LSTM(bi-directional long short-term memory)的网络入侵检测方法。所提方法通过采用引入梯度惩罚的GAN网络对样本不平衡问题进行处理，获取均衡的网络流量数据样本，并采用引入多层卷积层和注意力机制的Bi-LSTM网络对数据样本进行分类，实现了大数据环境下网络入侵检测。仿真结果表明，所提方法可有效检测模糊器、拒绝服务、漏洞利用等不同攻击类型的网络入侵，且具有较高的检测精度，准确率、精确率、召回率、F₁值分别达到90.28%、86.55%、93.27%、89.64%,假阳率为4.28%。相较于支持向量机(SVM)模型、随机森林(RF)模型和残差收缩网络(ResNet)模型，所提方法对大数据环境下网络入侵检测具有明显优势，为实现更精确的大数据环境下网络入侵检测提供了参考。

To improve the detection precision of network intrusion in the big data environment, a network intrusion detection method based on improved GAN(generative adversarial network) and improved Bi-LSTM(bi-directional long short-term memory) is proposed. The method adopts GAN network with gradient penalty to handle the problem of imbalanced samples, obtains balanced network traffic data samples, and uses Bi-LSTM network with multi-layer convolutional layers and attention mechanism to classify the data samples, achieving network intrusion detection in the big data environment. The simulation results show that the proposed method can effectively detect network intrusions of different types of attacks such as fuzzer, denial of service, and exploitation of vulnerability, which has higher detection precision. Moreover, its accuracy, precision, recall, and F₁ value reach 90.28%, 86.55%, 93.27%, and 89.64%, respectively, with a false positive rate of 4.28%. Compared with support vector machine(SVM), random forest(RF), and residual network(ResNet) models, the proposed method has significant advantages in network intrusion detection in the big data environment, providing a reference for achieving more accurate network intrusion detection in the big data environment.