Edited by: Amal Fawzi Ahmed , Ahmed Osama Ghazi , Mahmoud Ahmed Mahmoud , Muhammad Khaled Alsayed Atta , Muhammad Emad Aldin Nour Aldin ,Muhammad Kamal Suleiman
Download book version pdf
Evaluation of AI models to detect SQL injections in web applications
First edition “2025” – Book : Evaluation of AI models to detect SQL injections in web applications
All rights reserved to the #Democratic_Arabic_Center Germany – Berlin. Reproduction of this book or any part of it, or storing it in the scope of retrieving or transmitting the information in any form, without the prior Permission in writing of the publishe
Abstract
SQL injection attacks are still one of the most serious vulnerabilities targeting databases, requiring advanced technologies to detect them accurately and effectively. In this paper, we propose a hybrid model that combines BiLSTM’s serial capabilities with Transformer’s attention mechanisms for accurate detection of SQLi attacks. A special Tokenizer system based on SQL query structure analysis has been developed, along with the inclusion of an explanatory layer using SHAP to illustrate model decisions. Experiments with various data showed that the model outperformed only the CNN, BiLSTM, and Transformer models, achieving accuracy of 96.7%, and F1-score by 0.955. The results show that combining context and attention in a hybrid architecture represents a promising approach to securing databases, with interpretability High makes the model suitable for integration into production systems. SQL injection is an ongoing threat to databases, driven by unsecured query scripts. Deep learning techniques outperform traditional systems in detecting this type of attack, but unexplainable models are not suited to sensitive security contexts. In this paper, we present a hybrid model that analyzes query structure and contextualization via BiLSTM, and then enhances understanding via Transformer. We also use SHAP to provide a detailed interpretation of the classification, which enhances confidence in the system.
Publisher : Democratic Arabic Center For Strategic, Political & Economic Studies
