Optimizing neural networks for anomalous query detection in databases using the NSGA-II algorithm

Abstract Database security has become a critical challenge due to the rapid increase in malicious queries that attempt to exploit vulnerabilities in database systems. Traditional approaches, such as firewalls and access control lists, often fail to detect sophisticated and anomalous SQL queries. To address this issue, we propose a hybrid model that integrates a Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) architecture with the Non-dominated Sorting Genetic Algorithm II (NSGA-II) for optimized detection of anomalous queries. The CNN-LSTM network is employed to capture both spatial and temporal patterns within SQL queries. At the same time, NSGA-II is used to fine-tune hyperparameters, thereby improving classification accuracy and robustness. The model was validated using a standard SQL Injection dataset from the Kaggle repository, where it demonstrated superior performance compared to existing machine learning and deep learning approaches. This approach is particularly suited for deployment in high-security environments, such as banking and financial institutions, healthcare information systems, and enterprise-level applications, where protecting sensitive data against SQL injection attacks is of utmost importance. Experimental results confirm that the proposed method achieves higher accuracy, sensitivity, and reliability, making it a powerful tool for enhancing database security.