Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- asobirov/dga-preprocessed
|
| 5 |
+
- harpomaxx/dga-detection
|
| 6 |
+
- YangYang-Research/dga-detection
|
| 7 |
+
metrics:
|
| 8 |
+
- accuracy
|
| 9 |
+
- f1
|
| 10 |
+
tags:
|
| 11 |
+
- dga-detection
|
| 12 |
+
- streaming
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# DGA SGD Detector
|
| 16 |
+
|
| 17 |
+
This model was trained on the `asobirov/dga-preprocessed` dataset using:
|
| 18 |
+
- **HashingVectorizer** (char 2–5 grams, 2**20 hash buckets)
|
| 19 |
+
- **SGDClassifier** with online `partial_fit` and incremental learning-rate tuning
|
| 20 |
+
- Lightweight stats: length, digit ratio, Shannon entropy
|
| 21 |
+
- **Streaming training** on ~5 M domains, evaluated on ~1 M test domains from [harpomaxx/dga-detection](https://huggingface.co/harpomaxx/dga-detection) and [YangYang-Research/dga-detection](https://huggingface.co/YangYang-Research/dga-detection)
|
| 22 |
+
- **Decision threshold** optimized for F1 (=0.0794) yielding 85.7% accuracy, 83.6% DGA recall
|