Research On SQL Injection Detection Based On Deep Learning

The rapid development of Internet technology and the widespread of Web applications have brought a lot of convenience to people's daily life.At the same time,as more user information interacts between the browser and the server,attacks against Web applications intensify and network security issues occur frequently.Because of its simple implementation and high threat,structured query language(SQL)injection attack has always been a common tactic for cyber attackers.Nowadays,SQL injection attacks are endless,and the detection models based on traditional machine learning algorithms can not effectively identify complex SQL injection attacks.Therefore,this paper proposes to build a detection model based on deep learning algorithm to improve the detection effect.The main work of this paper is as follows:1.An improved convolutional neural network(CNN)classification model is proposed.A convolutional layer with a convolution kernel size of 2 and a step size of 2 is used instead of the pooling layer to solve the problem that pooling layer will lose low-level features and extract more data features while reducing dimensions.Using 1 * 1 convolutional layers instead of fully connected layers can greatly reduce the model parameters and improve model performance.2.An pre-training model of bidirectional encoder representation from transformers(BERT)and random forest algorithm are proposed to build SQL injection detection model.Through grid search and ten-fold cross-validation,some hyperparameters of the random forest algorithm are optimized to improve the detection accuracy.Based on the research of deep learning algorithm,this paper establishes SQL injection attack detection models and optimizes them from two perspectives of improving the classifier effect and text representation effect.The experimental results show that the detection model proposed in this paper is significantly more effective than that based on traditional machine learning algorithms,and improves the problem of insufficient detection accuracy of the current model.