基于卷积神经网络的跨站脚本攻击检测模型

Cross-site Scripting Attack Detection Model Based on Convolutional Neural Network

研究一种高效、准确的跨站攻击检测模型在信息安全领域中具有重要的意义。然而,传统的跨站攻击检测方法不仅需要花费大量的时间来提取这些攻击特征,还要结合一定的主观经验才能取得好的效果。为了提高XSS攻击的检测效率与准确率,论文提出一种基于一维卷积神经网络模型的XSS攻击检测方法。首先,根据XSS攻击样本的特点,对样本进行HTML与URL的解码与范化、分词以及向量化处理,然后将处理后的词向量输入到论文所设计的一维卷积神经网络模型中。通过多次实验选择模型合适的超参数,并与传统的检测模型进行对比实验。实验结果表明,论文提出的一维卷积神经网络模型以较少的检测时间达到了高达99.37%的准确率,与其它相关模型相比,此模型具有较好的性能与检测能力,对以XSS攻击为主的安全入侵检测与漏洞分析具有重要的意义。

It has important scientific significance and practical value to study an efficient and accurate cross-site attack(XSS)detection model in the field of information security.However,traditional XSS attack detection methods not only need to spend a lot of time to extract these attack features,but also combine certain subjective experience to achieve good results.To im⁃prove the detection efficiency and accuracy of XSS attacks,this paper proposes an XSS attack detection method based on a one-di⁃mensional convolutional neural network model.First,according to the characteristics of the XSS attack samples,the samples are de⁃coded and normalized by HTML and URL,word segmented and vectorized,and then the processed word vectors are input into the one-dimensional convolutional neural network model designed in this paper.The appropriate hyperparameters of the model are se⁃lected through multiple experiments,and a comparison experiment is performed with the traditional detection model.The experimen⁃tal results show that the one-dimensional convolutional neural network model proposed in this paper achieves an accuracy rate of up to 99.37%with less detection time.Compared with other related models,this model has better performance and detection ability,which is of great significance to XSS attack-based security intrusion detection and vulnerability analysis.