A HYBRID DETECTION FRAMEWORK FOR SQL INJECTION: REDUCING FALSE POSITIVES THROUGH SIMILARITY-BASED PATTERN MATCHING AND PROACTIVE DEVELOPER INTEGRATION
Abstract
SQL injection (SQLi) is a persistent and critical vulnerability in web applications that allows attackers to manipulate databases, access sensitive data, and disrupt services. Existing detection tools face challenges such as high false positive rates, limited effectiveness against obfuscated attacks, and poor integration into development workflows. This study proposes a hybrid SQLi detection framework that integrates similarity-based pattern matching with real-time feedback mechanisms embedded into Integrated Development Environments (IDEs) and Continuous Integration/Continuous Deployment (CI/CD) pipelines. The framework uses a novel algorithm that parses input queries into Abstract Syntax Trees (ASTs), employs fuzzy string matching to detect structural similarities with known attack patterns, and applies a predefined threshold mechanism to reduce false positives. Evaluation results indicate a detection accuracy of 97%, a false positive rate of 4%, and high developer usability, with an average Likert score of 4.6 out of 5. Compared to tools like OWASP ZAP and SQLMap, the framework demonstrates superior performance, particularly in handling complex and obfuscated SQLi patterns. Proactive integration into development workflows positions the framework as a solution not only for addressing SQLi vulnerabilities but also for enhancing secure coding practices, minimising costs, and supporting scalable adoption in DevSecOps-driven organisations. The framework sets a foundation for future advancements, including expanding detection capabilities to vulnerabilities like XSS and RCE and incorporating machine learning for adaptive and dynamic pattern analysis.