SQL Injection Attack Detection using Hybrid Model

SQL injection is among the worst type of threat because it lets the attacker manipulate the statements used in applications to corrupt the databases. It is proposed to design an intelligent real-time system for detecting and preventing the said type of attack using a combination of ML and DL approaches. Traditional machine learning techniques used in the detection part of the system include XGBoost and Random Forest while deep learning for the query analysis is done by Convolutional Neural Network (CNN). The application of extra features such as the TF-IDF and the character statistics makes it easier for the system to detect Phishing queries. The hybrid model that uses the output of both the ML and DL models has good prediction rates, and thus successfully mitigates SQL injection attacks with accuracy rate of 98.651%, precision rate of 99.688% and recall rate of 97.732%. The implemented system is based on the Flask web application allowing users to input queries and immediately get the results. This work shows how the application of machine learning and deep learning in and for cybersecurity, especially for SQL injection is highly useful and can provide a reliable solution to protect web applications today.