Detection of SQL Injection and Cross-Site Scripting Based on Multi-Model CNN Combined with Bidirectional GRU and Multi-Head Self-Attention

Both SQL injection and cross-site scripting (XSS) are critical threats in the field of web application and API security. According to the literatures and our experiments, use individual deep learning architecture such as DNN (Dense Neural Network), LSTM (Long short-term memory), RNN (Recurrent Neural Network) or transformer to detect SQL injection and XSS that requires much time and large amounts of data for training. For overcoming the above issues to balance the training performance and detection accuracy, we integrate CNN and GRU with the attention mechanism inspired by modern language models to construct an innovative and efficient detection system that can achieve higher accuracy, smaller training dataset and shorter training time. Moreover, the experimental results were even better than other baseline techniques and close to what we had anticipated.