SQL injection (SQLi) is a common security vulnerability for online applications that can lead to a database being compromised. Static criteria or signatures are used by traditional detection systems, but smart attackers can get around them. The SQL injection detection method that uses a neural network and Elasticsearch is more accurate and responsive. Our two-tiered design uses deep learning to sort SQL queries. Our convolutional neural network (CNN) picks out the details of SQL statements, finding patterns that other methods might overlook. Because it was trained on a wide range of conventional and dangerous SQL queries, the neural network may be able to apply what it learned to new types of attacks. Elasticsearch is used to search web application logs in real time on the second tier. By combining the output of a neural network with Elasticsearch, we can find SQL injection attacks faster. Compared to normal approaches, our hybrid method lowers the number of false positives and raises the number of detections. The model learns how to stop new SQL injection attacks and stops them from happening in the future. This framework protects web apps and can be used in more than one place. We use neural networks and powerful search methods to make SQL injection detection better. © 2025 Elsevier B.V., All rights reserved.