摘要
针对现有技术存在的网络攻击类别检测性能差、网络安全防御性能低等问题,开展基于大语言模型(LLM,Large Language Model)与Transformer机制的网络安全防御技术与应用研究。引入LLM与Transformer机制构建网络安全防御模型,采用LLM对构建模型训练数据进行采集与预处理;基于Transformer机制建立训练数据与网络攻击类别的映射关系,完成构建模型的预训练;以损失函数为依据,微调处理构建模型参数;根据构建模型预训练结果与构建模型参数微调结果,确定网络攻击类别,实现对网络攻击类别的有效检测与防御。实验结果表明,该技术的网络攻击类别检测效果优于对比技术,F1分数最大值可达0.9。
This paper focused on the problems of poor network attack category detection performance and low network security defense performance in existing technologies,and conducted research on network security defense technology and applications based on Large Language Model(LLM)and Transformer mechanism.It introduced LLM and Transformer mechanisms to construct a network security defense model,used LLM to collect and preprocess the training data of the constructed model,established a mapping relationship between training data and network attack categories based on the Transformer mechanism,and completed the pre-training of the model construction.Based on the loss function,the paper fined tune the parameters of the constructed model,and based on the pre-training results of the constructed model and the fine tuning results of the parameters,the paper determined the network attack category to implement effective detection and defense of the network attack category.The experimental results show that the network attack category detection effect of this technology is superior to the comparative technology,and the maximum F1 score can reach 0.9.
作者
杨伟青
周游
张婉澂
陈阳
YANG Weiqing;ZHOU You;ZHANG Wancheng;CHEN Yang(Institute of Information Technology,China Railway Shanghai Group Co.Ltd.,Shanghai 200071,China)
出处
《铁路计算机应用》
2025年第10期37-41,共5页
Railway Computer Application
