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 (
messagesfield) - 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:
{
"messages": [
{"role": "system", "content": "Arabic security assistant persona"},
{"role": "user", "content": "Arabic instruction + English security scenario"},
{"role": "assistant", "content": "Arabic security analysis"}
]
}
Usage
from datasets import load_dataset
dataset = load_dataset("haiderkamal23/allaM-offsec-arabic-chat")
License
Apache 2.0 - For authorized security research and testing only.