DeepURLBench: Multi-Class Dataset and DNS-Augmented Malicious-URL Classification
A cybersecurity contribution addressing malicious-URL detection. It provides DeepURLBench, a rigorously cleansed, open multi-class dataset distinguishing benign, phishing, and malicious URLs, and shows the multi-class framing outperforms binary classification for deep-learning models. It also augments the string-based URLNet classifier with DNS-derived features, improving accuracy while preserving real-time runtime efficiency.
2501.00356
This paper tackles malicious-URL classification for cybersecurity, addressing two gaps: the scarcity of comprehensive open-source datasets and the limitations of existing models that either lack real…