A novel deep learning technique for protecting web applications against XSS and SQLI attacks

The proliferation of emerging technologies has increased vulnerabilities in web applications, making them susceptible to SQL injections and cross-site scripting attacks. Current detection methods often yield high false positive rates. This paper introduces a new technique that improves web application security by decoding and standardizing SQL and XSS_payloads. A hybrid deep learning model was proposed, combining CNN for feature extraction and LSTM for sequential data dependencies. This model achieved impressive results, with accuracies of 99.79% for SQL-XSS Payload, 99.94% for Testbed, and 99.24% for HTTP CSIC 2010 datasets, demonstrating its effectiveness over traditional machine learning methods.