文献类型：期刊文献

英文题名：Deep Neural Network-Based SQL Injection Detection Method

作者：Zhang, Wei[1];Li, Yueqin[1];Li, Xiaofeng[1];Shao, Minggang[1];Mi, Yajie[1];Zhang, Hongli[1];Zhi, Guoqing[2]

通讯作者：Zhang, W[1];Li, YQ[1]

机构：[1]Beijing Union Univ, Smart City Coll, Beijing 100101, Peoples R China;[2]Beijing Union Univ, Coll Appl Arts & Sci, Beijing 100191, Peoples R China

第一机构：北京联合大学继续教育学院

通讯机构：[1]corresponding author), Beijing Union Univ, Smart City Coll, Beijing 100101, Peoples R China.|[1141733]北京联合大学继续教育学院;[11417]北京联合大学;

年份：2022

卷号：2022

外文期刊名：SECURITY AND COMMUNICATION NETWORKS

收录：;EI(收录号：20221611984799);Scopus(收录号：2-s2.0-85128256034);WOS:【SCI-EXPANDED(收录号:WOS:000793474400003)】；

语种：英文

外文关键词：Multilayer neural networks - Network security - Long short-term memory - Deep neural networks - Learning algorithms - Network layers

摘要：Among the network security problems, SQL injection is a common and challenging network attack means, which can cause inestimable loop-breaking and loss to the database, and how to detect SQL injection statements is one of the current research hotspots. Based on the data characteristics of SQL statements, a deep neural network-based SQL injection detection model and algorithm are built. The core method is to convert the data into word vector form by word pause method, then form a sparse matrix and pass it into the model for training, build a multihidden layer deep neural network model containing ReLU function, optimize the traditional loss function, and introduce Dropout method to improve the generalization ability of this model. The accuracy of the final model is maintained at over 96%. By comparing the experimental results with traditional machine learning algorithms and LSTM algorithms, the proposed algorithm effectively solves the problems of overfitting in machine learning and the need for manual screening to extract features, which greatly improves the accuracy of SQL injection detection.