haiderkamal23/allaM-offsec-arabic-chat

---
license: apache-2.0
task_categories:
- question-answering
- text-generation
language:
- ar
- en
tags:
- security
- offensive-security
- penetration-testing
- cybersecurity
- arabic
size_categories:
- 10K<n<100K
---

# Arabic Offensive Security Chat Dataset

Bilingual Arabic/English dataset for training offensive security assistants.

## Dataset Details

- **Training examples:** 18,412
- **Validation examples:** 2,000
- **Total:** 20,412
- **Languages:** Arabic (primary) + English (technical terms)
- **Format:** ChatML (`messages` field)
- **Source:** Filtered and processed from WNT3D/Ultimate-Offensive-Red-Team

## Intended Use

This dataset is designed for fine-tuning models to assist with:
- Vulnerability analysis and explanation
- Attack chain reasoning
- Penetration testing report generation
- OWASP/CWE/MITRE ATT&CK mapping
- Security mitigation recommendations

**IMPORTANT:** For authorized security testing and education only.

## Structure

Each example contains:
```json
{
"messages": [
{"role": "system", "content": "Arabic security assistant persona"},
{"role": "user", "content": "Arabic instruction + English security scenario"},
{"role": "assistant", "content": "Arabic security analysis"}
]
}
```

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("haiderkamal23/allaM-offsec-arabic-chat")
```

## License

Apache 2.0 - For authorized security research and testing only.