Enhancing SQL Injection Detection: CGAN with Discriminator Optimization Using Genetic Algorithm

SQL injection (SQLi) is the third most critical vulnerability in systems, posing significant threats to data and system integrity. Traditional detection methods often fail to adapt to evolving attack techniques, necessitating advanced solutions. This work utilizes Conditional Generative Adversarial Networks (CGAN) to generate realistic SQLi attacks and enhance detection systems, enabling adaptability to new attacker methods. More-over, the CGAN discriminator hyperparameters (i.e., neurons per layer, dropout rate, and number of layers) are fine-tuned using Genetic Algorithms (GA), which improved the accuracy and performance in detecting and classifying various SQLi attacks. The results show that this system outperforms traditional models like Naive Bayes, Decision Tree, Multi-layer Perceptron, Bagging (i.e., Random Forest), Stacking, and Boosting. This approach demonstrates strong capabilities in generating realistic attack scenarios, achieving data balance, and building robust SQLi defense systems for modern web applications.