---
configs:
- config_name: default
  data_files:
    - split: train
      path: "data/*.parquet"
---
<h1 align="center"> HiGraph: A Large-Scale Hierarchical Graph Dataset </h1>

<div align="center">

<img src="https://raw.githubusercontent.com/hzcheney/HiGraph/refs/heads/master/assets/h_logo.svg" width="100" height="100">

<p><strong>Hierarchical Graph Dataset for Malware Analysis with Function Call Graphs and Control Flow Graphs</strong></p>


<a href="https://arxiv.org/pdf/2509.02113"><img src="https://img.shields.io/badge/arXiv-2509.02113-b31b1b.svg" alt="arXiv" /></a>
                <img src="https://img.shields.io/badge/🤗%20Hugging%20Face-Datasets-yellow" alt="Hugging Face" />
                <img src="https://img.shields.io/badge/Dataset-6.17GB-blue" alt="Dataset Size" />
                <img src="https://img.shields.io/badge/Period-2012--2022-orange" alt="Time Period" />
                <img src="https://img.shields.io/badge/License-CC--BY--NC--SA-red" alt="License" />

A comprehensive hierarchical graph-based dataset for malware analysis and detection.

[Overview](#overview) • [Interactive Explorer](#interactive-visualization) • [Download](#download-dataset) 

</div>

## Overview

**HiGraph** is a novel, large-scale dataset that models each application as a hierarchical graph: a local **Control Flow Graph (CFG)** capturing intra-function logic and a global **Function Call Graph (FCG)** capturing inter-function interactions.

Graph-based methods have shown great promise in malware analysis, yet the lack of large-scale, hierarchical graph datasets limits further advances in this field. This hierarchical design facilitates the development of robust detection models that are more resilient to obfuscation, model aging, and malware evolution.


### Key Features

- 🔍 **Hierarchical Graph Structure**: Two-level representation with FCGs and CFGs
- 📈 **Large Scale**: 200M+ Control Flow Graphs and 499K+ Function Call Graphs
- 🏷️ **Rich Semantic Information**: Preserves crucial structural details for malware analysis
- 📊 **Comprehensive Coverage**: 11-year temporal span (2012-2022)
- 🎯 **Benchmark Ready**: Designed for advancing hierarchical graph learning in cybersecurity


## Interactive Visualization

Explore the hierarchical structure of malware samples through our interactive visualization tool:

<div align="center">

🔗 **[Launch Interactive Explorer](https://higraph.org/)**

*Click to explore the complete dataset structure and sample graphs*

</div>

## Download Dataset

Access the complete HiGraph dataset through multiple platforms:

<div align="center">

| Platform | Description | Link |
|----------|-------------|------|
| 🤗 **Hugging Face** | Primary dataset repository | [View on Hugging Face](https://huggingface.co/datasets/hzcheney/Hi-Graph/tree/main) |
| 🌐 **Project Page** | Interactive explorer | [HiGraph Explorer](https://higraph.org/) |

</div>

## Citation

If you find HiGraph useful in your research, please cite:

```
@article{chen2025higraph,
  title={HiGraph: A Large-Scale Hierarchical Graph Dataset for Malware Analysis},
  author={Chen, Han and Wang, Hanchen and Chen, Hongmei and Zhang, Ying and Qin, Lu and Zhang, Wenjie},
  journal={arXiv preprint arXiv:2509.02113},
  year={2025}
}
```

## Requirements
- Python >= 3.9
- torch==2.6.0
- torch-geometric==2.6.1

Install dependencies:

```bash
pip install -r requirements.txt
```

## 📄 License

This dataset is licensed under the **Creative Commons Attribution-NonCommercial-ShareAlike (CC-BY-NC-SA)** license. See the [LICENSE](LICENSE) file for details.