Answers are grounded in this paper and cite their sources.
This manuscript presents a character-level deep learning model that integrates a Convolutional Neural Network (CNN) layer with a Bidirectional Long Short-Term Memory (BiLSTM) layer to detect phishing URLs. The model is trained and evaluated using the PhiUSIIL Phishing URL Dataset, a publicly available benchmark recently released through the UCI Machine Learning Repository. The dataset comprises 235,795 URLs, including 134,850 legitimate URLs and 100,945 phishing URLs. URLs are tokenized at the character level and padded to a fixed length of 200 characters before being mapped to a 128-dimensional embedding space. The resulting representations are then passed through a one-dimensional convolutional (Conv1D) layer to extract local patterns, followed by max pooling and a Bidirectional LSTM layer to capture long-range sequential dependencies. Finally, a dense layer with a sigmoid activation function performs binary classification. On a held-out test set, the model achieves an accuracy of 97.8%, precision of 97.2%, recall of 98.1%, an F1-score of 97.6%, and a receiver operating characteristic area under the curve (ROC-AUC) of 0.987. Its performance is also compared with that of recent state-of-the-art phishing detection models reported in the literature from 2023 to 2026. Overall, this study contributes to the development of reliable and intelligent URL-based threat detection systems, helping to promote safer and more secure digital interactions.
No takes yet. Share an insight, caveat, or question.
A 2026 study studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: