This dataset is part of the ULE-CIBERLAB Project: Transfer of knowledge in cybersecurity for the country's business fabric, funded by the European Union NextGeneration-EU, Recovery, Transformation and Resilience Plan, through INCIBE.

MALWARE-SAMPLES DATASET

Disclaimer: This repository contains real samples of malware that can be executed (.exe) and artifacts related with their execution in CAPEv2 sandbox (JSON/HTML reports, screenshots, dropped files). DO NOT execute any of those binaries in production machines or non-isolated environments. This README file explains how dataset is structured, its metadata, safe use as well as legal and ethical requirements. This dataset has been created for research purposes only. The creators DO NOT accept responsibility for its use by third parties.

---

🌍 Available in other languages:

• Español

---

Content

1. Description
2. Source and attribution
3. Dataset structure
4. File description and format
5. CAPEv2 sandbox configuration
6. Contact

---

Description

This dataset only groups .exe samples extracted from three public repositories and their execution results in the CAPEv2 sandbox. For each sample, it includes:

• The binary (.exe)
• report.json generated by CAPEv2
• report.html (HTML version of thre report generated by CAPEv2)
• Screenshots taken by CAPE during the analysis
• Files dropped during dynamic execution (if any)

---

Source and attribution

The original binary samples were taken from public repositories (only .exe files were taken and used):

• ytisf/theZoo — https://github.com/ytisf/theZoo
• Pyran1/MalwareDatabase — https://github.com/Pyran1/MalwareDatabase
• Da2dalus/The-MALWARE-Repo — https://github.com/Da2dalus/The-MALWARE-Repo

---

Dataset structure

High-level structure (repository root):

dataset/
├── JSON/ 
│    └── <md5>.json # report.json (renamed using md5)
├── HTML/
│    └── <md5>.html # report.html (renamed using md5)
├── screenshots/
│    └── <md5>/     # md5 hash of the malware sample
│          └── md5.zip (Contains all image with names 0001.png, 0002.png...)
│          
├── binaries/
│    └── <md5>
|          ├── <md5>.exe
│          └── LICENSE.MD #Some binary files are under a GPLv3 license
└── dropped_files/
     └── <md5>/     # md5 hash of the malware sample
           └── md5.zip (Contains all files dropped in the system...)

---

File description and format

• binaries/<md5>/<md5>.exe
  Copy of the original executable file. DO NOT EXECUTE on uncontrolled computers.

• JSON/<md5>.json
  JSON output from CAPEv2 analysis: metadata, processes, dropped files, network activity, signatures, etc.

• HTML/<md5>.html
  HTML version of the report generated by CAPEv2.

• screenshots/<md5>/*.png
  Screenshots taken during dynamic execution.

• dropped_files/<md5>/*
  Files that the malware left on disk during its execution within the sandbox.

---

CAPEv2 sandbox configuration

Summary: In order to analyse samples with evasive capabilities, adjustments were made to the CAPEv2 sandbox configurations used during the generation of this dataset. In order to allow replication of the experiment, the configuration files used (the CAPEv2 conf/ folder) and the custom KVM virtual machine XML are included in the repository.

What does this section contain?

The following structure will be published within the repository:

CAPEv2 Configuration/
├── XML KVM/
│     └── xml-custom-machine.xml # XML used to configure the VM in KVM
└── conf/
      └── ... # CAPEv2 configuration files and folders

• XML KVM/
  Contains the XML descriptor of the virtual machine (suggested names: xml-custom-machine.xml). This XML reflects the VM used in the analyses (vCPU, memory, virtual devices, disks, simulated NICs, etc.) so that others can import/define an equivalent VM in their virtualisation environment.

• conf/
  Copy of the conf folder used by CAPEv2 in the experiment. It includes the relevant CAPEv2 configuration files (paths, timeouts, enabled modules, integration with the analysis backend, network capture parameters, etc.). Note: any sensitive data (credentials, private keys) has been removed.

---

How to Reference

If you use this model in your research or project, please reference it as:

Bayón-Martínez, R., & Prieto-González, A. (2025). Malware Samples Dataset [Data set]. Grupo de Robótica de la Universidad de León. Hugging Face.
https://huggingface.co/datasets/unileon-robotics/malware-samples
License: Creative Commons BY-NC-SA 4.0.

You can also make use of this BibTeX extract:

@dataset{bayonmartinez_prietogonzalez_2025_malware_samples,
  author       = {Bayón-Martínez, Raúl and Prieto-González, Adrian},
  title        = {Malware Samples Dataset},
  year         = {2025},
  publisher    = {Grupo de Robótica de la Universidad de León},
  howpublished = {\url{https://huggingface.co/datasets/unileon-robotics/malware-samples}},
  note         = {Available on Hugging Face. License: CC BY-NC-SA 4.0}
}

---

Contact

Raúl Bayón Martínez

PhD Student/Research Staff from Universidad de León (Spain)

Email: rbaym@unileon.es

---

Adrián Prieto González

Research Staff from Universidad de León (Spain)

Email: aprig@unileon.es