Enhancing Web Application Security Through Automated SQL Injection Detection Using Neural Networks

Abstract

SQL injection (SQLi) attacks are one of the major threats to web application security, especially on e-commerce platforms. These attacks exploits the weaknesses in user input which enables attackers to have access and manipulate database queries to compromise data integrity. This study aims to develop an automated SQL injection detection system using Neural Networks to improve on the security of web applications. A labeled dataset of SQL injection patterns was created to train three machine learning models namely: Naive Bayes, Random Forest, and Deep Neural Network. The models evaluation was done using accuracy, precision, specificity, and F1 score. The results shows that the Neural Network model outperformed the two others by achieving an accuracy of 99.1%, a precision of 94.2%, a specificity of 98.1%, and a F1 score of 0.961. These results shows that Neural Networks is efficient in detecting SQL injection attacks. Finally, this study provides a comparative insights to earlier research by exploring different potential deployment scenarios, and identifies avenues for future work.