Upload model_card.md with huggingface_hub
Browse files- model_card.md +38 -0
model_card.md
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: en
|
| 3 |
+
license: other
|
| 4 |
+
tags:
|
| 5 |
+
- security
|
| 6 |
+
- phishing-detection
|
| 7 |
+
- url-classification
|
| 8 |
+
- xgboost
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# Random Forest / XGBoost Model for URL Phishing Detection
|
| 12 |
+
|
| 13 |
+
## Model Details
|
| 14 |
+
- Architecture: Gradient-boosted decision trees (XGBoost)
|
| 15 |
+
- Input: Single URL string (no external queries)
|
| 16 |
+
- Features: Lexical and structural URL features (lengths, symbol counts, digit ratio, IPv4 pattern, common phishing tokens, scheme/TLD heuristics)
|
| 17 |
+
- Training data: `PhiUSIIL_Phishing_URL_Dataset.csv`
|
| 18 |
+
- Intended use: Binary classification (phishing vs. legitimate)
|
| 19 |
+
|
| 20 |
+
## Metrics (test)
|
| 21 |
+
- Accuracy: 0.9952
|
| 22 |
+
- Precision: 0.9928
|
| 23 |
+
- Recall: 0.9989
|
| 24 |
+
- F1: 0.9958
|
| 25 |
+
- ROC-AUC: 0.9976
|
| 26 |
+
|
| 27 |
+
## Usage
|
| 28 |
+
See `README.md` and `inference.py` for loading and `predict_url()`.
|
| 29 |
+
|
| 30 |
+
## Limitations and Biases
|
| 31 |
+
- URL-only features can be evaded by sophisticated attackers.
|
| 32 |
+
- Dataset shifts and novel TLDs may degrade performance.
|
| 33 |
+
- Always validate on your own traffic before deployment.
|
| 34 |
+
|
| 35 |
+
## License
|
| 36 |
+
Provided for research/educational purposes. Ensure compliance with local laws and organizational policies.
|
| 37 |
+
|
| 38 |
+
|