.png){kind=icon}
.png){kind=icon}
Return
MLF-ICL: Adaptive malicious URL detection via multi-level feature fusion and TabDPT-based in-context learning
.png){kind=icon}
OA
AI
L
F
W
K
DOI：10.1016/j.neucom.2026.134755.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Enhanced DCG-BERT uses Multi-order Dynamic Feature Fusion to link character anomalies with token intent in obfuscated URLs. • Deep semantic embeddings and Saliency Map-filtered lexical rules unite implicit data-driven patterns with expert knowledge. • We pioneer TabDPT with ICL, shifting static parameter fitting to retrieval-augmented adaptive inference for zero-day threats. • Tests achieve state-of-the-art F1-score (99.78%) on highly imbalanced data, outperforming CNN/LSTM under concept drift.
Keywords:
Malicious URL detection
Multi-view feature fusion
Lexical analysis features
Tabular foundation model
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
6.5
Papers:
2.5W
Citations:
6.5W
Organization
Cited Papers