|
|
--- |
|
|
license: mit |
|
|
tags: |
|
|
- cybersecurity |
|
|
- africa |
|
|
- threat-detection |
|
|
- NLP |
|
|
- Allsafeafrica |
|
|
- cyber-aware |
|
|
datasets: |
|
|
- HuggingFaceFW/fineweb-2 |
|
|
metrics: |
|
|
- accuracy |
|
|
- bertscore |
|
|
base_model: |
|
|
- HuggingFaceTB/SmolLM3-3B |
|
|
- google/gemma-3n-E4B-it |
|
|
new_version: HuggingFaceTB/SmolLM3-3B |
|
|
pipeline_tag: text-classification |
|
|
library_name: adapter-transformers |
|
|
--- |
|
|
|
|
|
# 🛡️ Cyber Threat Detector Africa |
|
|
|
|
|
> Developed by [Allsafeafrica](https://huggingface.co/allsafeafrica) |
|
|
> A lightweight NLP model built to detect and classify potential cybersecurity threats in textual data across African SMEs, startups, and digital communities. |
|
|
|
|
|
--- |
|
|
|
|
|
## 📌 Overview |
|
|
|
|
|
**Cyber Threat Detector Africa** is an AI-powered model designed to: |
|
|
- Classify cyber risk indicators in natural language (emails, messages, reports) |
|
|
- Support awareness in employee training platforms |
|
|
- Act as a backend tool for ESG-cyber hybrid security assessments |
|
|
|
|
|
--- |
|
|
|
|
|
## 🧠 Model Info |
|
|
|
|
|
| Attribute | Detail | |
|
|
|----------|--------| |
|
|
| Framework | `transformers`, `pytorch` | |
|
|
| Base Model | `distilbert-base-uncased` | |
|
|
| Fine-tuned On | Synthetic + local African threat incident data | |
|
|
| Labels | `phishing`, `malware`, `social-engineering`, `safe`, `suspicious` | |
|
|
| Accuracy | ~91.7% on test set | |
|
|
|
|
|
--- |
|
|
|
|
|
## ✨ Example Usage |
|
|
|
|
|
```python |
|
|
from transformers import pipeline |
|
|
|
|
|
threat_detector = pipeline("text-classification", model="allsafeafrica/cyber-threat-detector-africa") |
|
|
|
|
|
text = "Your account has been suspended. Click here to verify your identity." |
|
|
threat_detector(text) |
