基于不平衡数据的网络流量异常检测方法研究  

Research on abnormal detection method of network traffic based on unbalanced data

作　　者：蔡登江 CAI Dengjiang(China National Offshore Oil Corporation Information Technology Center,Beijing 100010,China)

机构地区：[1]中国海洋石油集团有限公司信息技术中心,北京100010

出　　处：《电子设计工程》2025年第1期46-50,共5页Electronic Design Engineering

摘　　要：为有效解决不平衡数据影响的问题,确保面对大规模网络流量数据异常检测的实时性,提出了基于不平衡数据的网络流量异常检测方法。通过优化SMOTE(合成少数类过采样)算法对含不平衡数据的网络流量数据进行平衡处理,将得到的数据集通过核主成分分析方法实现特征提取后,输入到卷积神经网络中。通过卷积和池化过程进一步实现网络流量数据深度特征提取,依据Softmax分类层对网络流量特征进行分类,利用训练好的卷积神经网络预测模型实现不平衡数据的网络流量异常检测。通过实验验证,该方法展现出了良好的效率和稳定性。在迭代次数为40次时,实现最佳不平衡数据处理结果,能够对异常数据进行精准识别。To effectively address the impact of imbalanced data and ensure real⁃time detection of anomalies in large⁃scale network traffic data,a network traffic anomaly detection method based on imbalanced data is proposed.The network traffic data containing imbalanced data is balanced by optimizing the SMOTE(Synthetic Minority Oversampling Technique)algorithm.The obtained dataset is then feature extracted using kernel principal component analysis and input into a convolutional neural network.Further achieve deep feature extraction of network traffic data through convolution and pooling processes,classify network traffic features based on the Softmax classification layer,and use the trained convolutional neural network prediction model to achieve network traffic anomaly detection of imbalanced data.Through experimental verification,this method demonstrates good efficiency and stability.When the number of iterations is 40,achieve the best unbalanced data processing result,and ultimately be able to accurately identify abnormal data.

关 键 词：不平衡数据 网络流量异常检测 优化SMOTE算法 核主成分分析 卷积神经网络 Softmax分类 

分 类 号：TN915.08[电子电信—通信与信息系统]