Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
tags:
|
| 6 |
+
- cybersecurity
|
| 7 |
+
- vulnerability
|
| 8 |
+
- mitre-attck
|
| 9 |
+
- text-classification
|
| 10 |
+
- fine-tuned
|
| 11 |
+
base_model: ehsanaghaei/SecureBERT
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
# SecureBERT — MITRE ATT&CK Classifier
|
| 15 |
+
|
| 16 |
+
Part of the **CVE-LMTune** model suite — language models fine-tuned for multi-taxonomy vulnerability classification.
|
| 17 |
+
|
| 18 |
+
## Paper
|
| 19 |
+
|
| 20 |
+
> Franco Terranova, Sana Rekbi, Abdelkader Lahmadi, Isabelle Chrisment. *Multi-Taxonomy Vulnerability Classification with Hierarchically Finetuned Language Models.* DIMVA '26.
|
| 21 |
+
|
| 22 |
+
[](https://opensource.org/licenses/MIT)
|
| 23 |
+
[](https://zenodo.org/records/17368476)
|
| 24 |
+
|
| 25 |
+
## Task
|
| 26 |
+
|
| 27 |
+
**MITRE ATT&CK technique classification from CVE descriptions**
|
| 28 |
+
|
| 29 |
+
## Performance
|
| 30 |
+
|
| 31 |
+
See paper for details
|
| 32 |
+
|
| 33 |
+
## Model Structure
|
| 34 |
+
|
| 35 |
+
flat — standard `AutoModelForSequenceClassification`
|
| 36 |
+
|
| 37 |
+
## Quick Start
|
| 38 |
+
|
| 39 |
+
```python
|
| 40 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
| 41 |
+
import torch
|
| 42 |
+
|
| 43 |
+
tokenizer = AutoTokenizer.from_pretrained("Sana9/securebert-mitre-attack")
|
| 44 |
+
model = AutoModelForSequenceClassification.from_pretrained("Sana9/securebert-mitre-attack")
|
| 45 |
+
model.eval()
|
| 46 |
+
|
| 47 |
+
text = "Buffer overflow vulnerability in OpenSSL allows remote attackers to execute arbitrary code."
|
| 48 |
+
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
|
| 49 |
+
|
| 50 |
+
with torch.no_grad():
|
| 51 |
+
logits = model(**inputs).logits
|
| 52 |
+
probs = torch.sigmoid(logits) # multi-label → sigmoid
|
| 53 |
+
```
|
| 54 |
+
|
| 55 |
+
> **Note for hierarchical models:** This repo contains multiple sub-folders (master + slave models).
|
| 56 |
+
> Load each sub-folder separately using `from_pretrained("Sana9/securebert-mitre-attack/master")` etc.
|
| 57 |
+
|
| 58 |
+
## Citation
|
| 59 |
+
|
| 60 |
+
```bibtex
|
| 61 |
+
@inproceedings{terranova2026cvelmtune,
|
| 62 |
+
title = {Multi-Taxonomy Vulnerability Classification with Hierarchically Finetuned Language Models},
|
| 63 |
+
author = {Terranova, Franco and Rekbi, Sana and Lahmadi, Abdelkader and Chrisment, Isabelle},
|
| 64 |
+
booktitle = {Proceedings of DIMVA '26},
|
| 65 |
+
year = {2026}
|
| 66 |
+
}
|
| 67 |
+
```
|
| 68 |
+
|
| 69 |
+
## Related Resources
|
| 70 |
+
|
| 71 |
+
- 🤗 [Full model suite](https://huggingface.co/Sana9)
|
| 72 |
+
- 📦 [Zenodo data & logs](https://zenodo.org/records/17368476)
|