Phishing URL recognition based on ON-LSTM attention mechanism and XGBoost model
Qiaojuan Jia, XingKe Guo, Ming Zhang, Ming Liu, XiaoYun Tian, XiaoYu Jin, DaWei Ma · 2023
Aiming at the problems of losing sensitive words and low detection performance in phishing URL recognition, a phishing URL recognition method based on ON-LSTM attention mechanism and XGBoost model is proposed. Firstly, the URL is divided according to the word level of sensitive words, and the hierarchical features of URL are fully extracted through the unsupervised self-learning feature of the URL by the ON-LSTM mode. Then, the attention mechanism is used to assign weights to the feature vectors after feature extraction, so as to further improve the depth of URL feature extraction. Finally, the XGBoost classification model is constructed, the XGBoost model is trained by feature vectors to realize URL classification and recognition, which effectively improves the performance and efficiency of model classification. The experimental results show that compared with the comparison model, this method has better ability and performance to identify phishing URLs.