Enhancing SQL injection detection: a comparative study of SVM, Random Forest and hybrid models

Web applications are still at serious risk from SQL injection (SQLi) attacks, which take advantage of lax user input validation. In order to identify fraudulent SQL queries, this study contrasts Support Vector Machines (SVM), Random Forests (RF), and a hybrid SVM-RF model. To direct the creation of reliable SQLi detection systems, we evaluate classification accuracy, generalization to invisible threats, computational efficiency, and interpretability. We expanded the study to improve novelty by incorporating deep learning models that are well-suited to capture the sequential structure of SQL queries: CNN-BiLSTM hybrid, bidirectional LSTM (BiLSTM), and long short-term memory (LSTM). Superior detection performance against intricate and obfuscated injection patterns was shown by these models. Future research will concentrate on enhancing detection through enhanced feature engineering and deep learning advancements.