Research on malicious domain name detection method based on deep learning

This study aims to address complex network attacks by proposing a malicious domain detection model based on a combination of BERT-TextCNN with attention mechanisms. The BERT model is employed to learn contextual semantics and generate rich semantic representations, while TextCNN contributes local feature extraction capabilities. The integration of global and local attention mechanisms allows targeted focus on key information in URLs, enhancing adaptability to various attack methods. Experimental results across multiple datasets demonstrate the model's outstanding performance in accuracy, precision, recall, and F1-Score, achieving an accuracy rate of 96.67%. In comparison to traditional methods, the proposed model maintains high detection accuracy while exhibiting a broader detection range, providing a reliable deep learning solution for malicious domain detection.