摘要
针对传统电网网络恶意流量检测算法准确率低、训练效率差的缺点,文中利用不同粒度大小的对象集提出一种网络安全指标判别模型。该方案将数据包、网络流以及会话流量化为数学模型,使用CNN构建通道注意力和空间注意力机制,提取输入数据的空间特征,使用轻量化双向GRU模型提取数据的时序特征,并将这两种特征进行融合,利用信息熵对网络安全权重进行测算,得到流量分类结果。算法模型的消融实验测试结果表明,算法改进是有效的,对比实验中所提算法准确率、召回率和F1值均为最优,运行时间在100 s以内,证明了该算法兼具性能和效率。
In response to the shortcomings of low accuracy and poor training efficiency of traditional malicious traffic detection algorithms in power grid networks,this paper proposes a network security indicator discrimination model using object sets of different granularity sizes.This scheme converts data packets,network flows,and session traffic into mathematical models,uses CNN to construct channel attention and spatial attention mechanisms,extracts spatial features of input data,uses lightweight bidirectional GRU model to extract temporal features of data,and combines these two features.Information entropy is used to calculate network security weights,and finally obtains traffic classification results.The ablation experimental results of the algorithm model show that the algorithm improvement is effective.The accuracy,recall,and F1 values of the proposed algorithm in the comparative experiment are all optimal,and the running time is within 100 seconds,proving that the algorithm has both performance and efficiency.
作者
李昂
钱锦
陈元中
徐汉麟
罗俊
LI Ang;QIAN Jin;CHEN Yuanzhong;XU Hanlin;LUO Jun(Hangzhou Power Supply Company,State Grid Zhejiang Electric Power Co.,Ltd.,Hangzhou 310000,China)
出处
《电子设计工程》
2025年第7期140-144,共5页
Electronic Design Engineering
基金
国家电网公司总部科技项目(5700-202319297A-1-1-ZN)
国网浙江省电力有限公司科技项目(5211XT22000W)。
关键词
对象集选择
恶意流量检测
注意力机制
GRU模型
信息熵
电网网络安全
object set selection
malicious traffic detection
attention mechanism
GRU model
information entropy
grid network security
作者简介
李昂(1988-),男,江苏连云港人,硕士,工程师。研究方向:信息安全。
