Malicious URL detection and research based on deep learning
In the digital era,cybersecurity issues are particularly prominent,with the widespread dissemination of malicious URLs posing a significant threat to personal privacy and enterprise security.While research in English-language environments has progressed,studies on Chinese online environments remain relatively scarce,and there is a lack of large-scale Chinese URL datasets.To address this gap,this research constructed a large-scale Chinese online URL dataset and proposed a hybrid model(BiAC)based on Bidirectional Long Short-Term Memory Networks(BiLSTM)and Attention-augmented Convolutional Neural Networks(CNN)for detecting malicious URLs.The BiAC model significantly enhanced detection accuracy and efficiency by deeply integrating the temporal and grammatical features captured by BiLSTM with the semantic features extracted by Attention-augmented CNN.Experimental results showed that the BiAC model achieved a remarkable accuracy of 97.53％and an F1 Score of 93.05％on the task of detecting malicious Chinese online URLs,surpassing traditional models.This achievement not only validated the effectiveness of the model design but also highlighted the importance of constructing datasets specifically tailored for Chinese environments.It has significant practical implications and application value for enhancing cybersecurity protection capabilities.
deep learningmalicious URL detectionconvolutional neural networks(CNN)
NETL
NSTL
万方数据
