HYBRID FRAMEWORKS FOR DETECTING AND PREVENTING SQL INJECTION ATTACKS IN DATABASE-DRIVEN APPLICATIONS
Keywords:
SQL injection attack(SQLIA), database/SQL queries, web applications, Blind SQL, XSS attack, Deep Learning, Graph Convolutional Network (GCN), LSTM, FastText.Abstract
SQL Injection Attacks (SQLIAs) remain a major treat to database-driven applications, enabling data theft, authentication bypass, and system disruption. This study evaluates common SQLi techniques (Union Error-based, Blind) using penetration-testing tools such as Burp Suite, and reviews existing defenses including input validation, parameterized queries, and anomaly detection. We suggest a hybrid deep learning framework that combines LSTM, Graph Convolutional Networks (GCN), and FastText embeddings for better SQLIA detection and prevention in order to get around the drawbacks of traditional rule-based systems. We highlight their strengths and limitations, and propose improved mitigation strategies through a structured analysis of attack vectors and countermeasures. The findings provide practical guidance for developers and security professionals to strengthen application resilience against SQLIAs.
