Research on Injection Attack Detection Method Based on Bert-CNN

Traditional injection attack detection methods suffer from low accuracy and limited effectiveness against web attacks. To address the challenges of low detection accuracy and complex data features in SQL injection attack detection, this paper proposes a Bert-CNN-based detection model. By integrating Bert's deep semantic understanding with the convolutional neural network's local feature extraction capability, the model effectively identifies and classifies normal and malicious SQL query requests, enhancing the security of critical applications such as power information systems. Experimental results show that, compared to SQL injection detection methods based on CNN, LSTM, and Bert-Base, the proposed model achieves improvements in accuracy, precision, recall, and F1-score on the test set, with faster convergence of the target loss function, demonstrating higher detection efficiency.