基于DCNN-Transformer模型的XSS攻击检测方法  

XSS attack detection method based on DCNN-Transformer model

作　　者：何志伟 高大鹏[1] HE Zhi-wei;GAO Da-peng(Civil Aviation Flight University of China,Guanghan 618307,Sichuan Province,China)

机构地区：[1]中国民用航空飞行学院计算机学院,四川广汉618307

出　　处：《信息技术》2025年第3期93-100,共8页Information Technology

基　　金：飞行学院智慧民航专项重点科研项目(ZHMH2022-002);飞行学院2020年度科研面上项目(J2020-064)。

摘　　要：为进一步提高XSS攻击的检测效果,文中提出一种基于DCNN-Transformer模型的XSS攻击检测方法。通过对收集的数据依次进行解码、规范化、分词、TF-IDF选词、构建词典和编码预处理,用于模型的训练和测试。文中提出的DCNN-Transformer模型引入了Embedding层,还综合了一维深度卷积神经网络快速处理序列数据和Transformer模型并行处理序列数据及学习序列元素间依赖关系的能力。实验结果表明,DCNN-Transformer模型相比于LSTM、GRU、DCNN和DCNN-GRU模型,收敛速度最快且效果更优,准确率、召回率和f1值最高,模型轻量、检测速度快,综合表现显著优于其他4个模型,为XSS攻击检测提供了一个更优的方法。In order to further improve the detection effect of XSS attack,an XSS attack detection method based on DCNN-Transformer model is proposed.Through the decoding,normalization,word segmentation,TF-IDF word selection,dictionary construction and coding preprocessing of the collected data,it is used for model training and testing.The DCNN-Transformer model proposed in this paper introduces the Embedding layer,and also integrates the ability of one-dimensional deep convolutional neural network to quickly process sequence data and the Transformer model to parallel process sequence data and learn the dependencies between sequence elements.The experiment results show that the DCNN-Transformer model has the fastest convergence speed and better effect than the LSTM,GRU,DCNN and DCNN-GRU models.The accuracy,recall rate and f1 value are the highest.The model is lightweight and the detection speed is fast.The comprehensive performance is significantly better than the other four models,which provides a better method for XSS attack detection.

关 键 词：XSS攻击检测 卷积神经网络 Transformer Embedding层 TF-IDF 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]