haiderkamal23/allaM-offsec-arabic-chat-v2

---
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 v2

**High-quality category-aware bilingual Arabic/English dataset for offensive security assistants.**

## What's New in v2

- ✅ **Category-aware responses:** Different response structures for web vulns, DeFi, reconnaissance tools, social engineering, etc.
- ✅ **No generic templates:** Each category has specialized analysis framework
- ✅ **No verbatim copying:** Responses analyze and transform the input, not repeat it
- ✅ **Semantic accuracy:** Tools (nmap, recon-ng) described as tools, not vulnerabilities
- ✅ **Real attack chain reasoning:** Detailed multi-stage attack scenarios

## Dataset Details

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

## Categories Covered

- **general:** 8086 examples
- **web_vuln:** 3753 examples
- **defi_crypto:** 2042 examples
- **recon_tools:** 1890 examples
- **malware_ransomware:** 1814 examples
- **api_protocol:** 994 examples
- **infra_network:** 885 examples
- **cloud_container:** 858 examples
- **social_engineering:** 90 examples

## Response Structures

### Web Vulnerabilities
- Vulnerability type identification
- Technical analysis
- Multi-stage attack chain
- OWASP/CWE/MITRE mapping
- Specific mitigations (not generic)
- Detection techniques

### DeFi/Smart Contracts
- Blockchain-specific attack vectors
- Reentrancy, flash loans, oracle manipulation
- On-chain impact analysis
- Solidity-specific mitigations
- Audit tools and formal verification

### Reconnaissance Tools
- Tool descriptions (not vulnerabilities!)
- Role in attack chain
- Detection from defender perspective
- Network monitoring strategies

### Social Engineering
- Attack psychology and techniques
- Delivery mechanisms
- User awareness as primary defense
- Email/communication security

### Malware/Ransomware
- Full kill chain (10+ stages)
- Lateral movement tactics
- Double extortion analysis
- Backup and recovery strategies

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("haiderkamal23/allaM-offsec-arabic-chat-v2")
```

## License

Apache 2.0 - For authorized security research and testing only.

## Training

Designed for QLoRA fine-tuning of:
- `humain-ai/ALLaM-7B-Instruct-preview`
- Or similar Arabic LLMs

Target model: `haiderkamal23/ALLaM-7B-OffSec-Arabic-v2`