An Attention Module Integrated Deep Learning Architecture with BERT to Detect SQL Injection Attacks

SQL injection attacks rank among the most prevalent forms of attacks targeting web applications. These attacks capitalize on weaknesses within an application's database access protocols, enabling malicious actors to carry out unauthorized SQL queries. Therefore, detection of these types of attacks is considered as one of the important research areas. Currently, machine learning-based methods have gained huge attention due in this domain because of facilitating automated approach for attack detection. However, attack detection accuracy remains a challenging task. To overcome this issue, we adopt Deep Learning mechanism and present a new architecture by using BiLSTM, and attention mechanism which helps to extract the fine contextual information. The performance of this approach is validated on SQLIA dataset where it reported the overall performance as 98.15, 99.20, 99.15, and 98.50% in terms of accuracy, precision, recall, and f1-score respectively.