README.md

---
tags:
- ml-intern
---
# VulnAI / VulnIA - Java Vulnerability Detection Dataset V3

## Purpose
This dataset is designed exclusively for **defensive** Java vulnerability detection using CodeBERT fine-tuning.

## Task
Multi-label classification of Java source code snippets into vulnerability families and fine-grained sub-types.

## Labels
- 0-65: V1 legacy labels (preserved, not modified)
- 3: XXE_GENERAL (V1 replay)
- 14: DESER_GENERAL (V1 replay)
- 66: CRYPTO
- 67: SAFE
- 68-78: Fine-grained XXE labels
- 79-88: Fine-grained Deserialization labels

Total labels: 89

## Parent / Fine-grained relationship
- Fine labels 68-78 → parent 3 (XXE)
- Fine labels 79-88 → parent 14 (DESER)
- Labels 66, 67 → parent = self

## Splits
- train: 1181
- validation: 238
- test: 236
- total: 1655

## Families
{'deserialization': 580, 'xxe': 575, 'crypto': 300, 'safe': 200}

## Preprocessing
- Anti-leakage cleaning (method/class names, comments)
- Exact deduplication (SHA-256 on normalized code)
- Near-duplicate grouping (n-gram Jaccard ≥ 0.85)
- Token-length handling (max 512 tokens for CodeBERT)
- Group-based split to prevent leakage

## Ethical Use
- This dataset is for **defensive security research only**.
- Do NOT use to generate exploits, payloads, or malware.
- All payloads and secrets have been removed or neutralized.

## License
Apache-2.0

<!-- ml-intern-provenance -->
## Generated by ML Intern

This dataset repository was generated by [ML Intern](https://github.com/huggingface/ml-intern), an agent for machine learning research and development on the Hugging Face Hub.

- Try ML Intern: https://smolagents-ml-intern.hf.space
- Source code: https://github.com/huggingface/ml-intern

## Usage

```python
from datasets import load_dataset

dataset = load_dataset('MaryamEl/vulnai-java-v3-dataset')
```