|
| 基于混合自编码器模型的XSS攻击检测研究 |
| Research on XSS Attack Detection Based on Hybrid Autoencoder Models |
| 投稿时间：2025-12-24 |
| DOI：10.16018/j.cnki.cn32-1650/n.202601003 |
| 中文关键词: XSS攻击检测 自编码器 BERT USE |
| 英文关键词: XSS attack detection autoencoder BERT USE |
| 基金项目:滁州市委网信办横向项目(HX2022122) |
|
| 摘要点击次数: 3 |
| 全文下载次数: 1 |
| 中文摘要: |
| 针对传统检测方法,包括基于规则的方法和传统深度学习模型难以适应新兴攻击技术,难以有效捕捉跨站脚本(Cross-site scripting, XSS)攻击载荷细微差异与上下文相关性的局限,提出了一种融合检测模型UBL-XSS。该模型创新性地结合了BERT、通用句子编码器和基于LSTM的自编码器3种深度学习技术,从多个维度捕获XSS攻击载荷的语义和句法特征,实现了基准数据集上99.95%的精确度和99.84%的召回率,显著优于文中提出的现有主要检测方法。 |
| 英文摘要: |
| To address the limitations of traditional detection methods, including rule-based approaches and conventional deep learning models, which struggle to adapt to emerging attack techniques and fail to effectively capture the subtle differences and contextual dependencies of cross-site scripting(XSS) payloads, this paper proposes a fusion detection model named UBL-XSS. The model innovatively integrates three deep learning techniques including BERT, universal sentence encoder, and LSTM-based autoencoders, to capture the semantic and syntactic features of XSS payloads from multiple dimensions. Experimental results demonstrate that UBL-XSS achieves a precision of 99.95% and a recall of 99.84% on benchmark datasets, significantly outperforming the state-of-the-art detection methods discussed in this study. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|