README.md

---
license: cc-by-4.0
size_categories:
- 1K<n<10K
task_categories:
- image-classification
- image-segmentation
configs:
- config_name: 1024-unrolled
  data_files:
  - split: train
    path: 1024-unrolled/*train*.tar
  - split: valid
    path: 1024-unrolled/*valid*.tar
  - split: test
    path: 1024-unrolled/*test*.tar
- config_name: 2048-unrolled
  data_files:
  - split: train
    path: 2048-unrolled/*train*.tar
  - split: valid
    path: 2048-unrolled/*valid*.tar
  - split: test
    path: 2048-unrolled/*test*.tar
- config_name: 4096-unrolled_n
  data_files:
  - split: train
    path: 4096-unrolled_n/*train*.tar
  - split: valid
    path: 4096-unrolled_n/*valid*.tar
  - split: test
    path: 4096-unrolled_n/*test*.tar
---


# COSOCO: Compromised Software Containers Image Dataset

- **Paper:** [Malware Detection in Docker Containers: An Image is Worth a Thousand Logs](https://huggingface.co/papers/2504.03238)
- **Dataset Documentation:** [COSOCO Dataset Documentation](./docs/COSOCO-dataset-readme-v1_0.pdf)

## Dataset Description

COSOCO (Compromised Software Containers) is a synthetic dataset of 3364 images representing benign
and malware-compromised software containers. Each image in the dataset represents a dockerized 
software container that has been converted to an image using common byte-to-pixel tools widely used
in malware analysis. Software container records are labelled (1) **benign** or (2) **compromised**: 
A benign software container will have installed commonly used harmless packages and tools, whereas 
a compromised software container, will have, among harmless benign tools and packages, its underlying
file system affected by some activated malware instance. Each compromised instance is accompanied by 
a mask, i.e. a black and white image which marks the pixels that correspond to the files of the 
underlying system that have been altered by a malware. COSOCO aims to support the identification of 
compromised software containers via the task of image classification task and the identification of 
compromised files and file system regions inside a container via the image segmentation task.

## Acknowledgements

This project has received funding from the European Union’s Horizon Europe research and innovation
programme under grant agreement **No 101093069 (P2CODE)**. Disclaimer: Funded by the European Union. Views
and opinions expressed are however those of the author(s) only and do not necessarily reflect those of
the European Union or European Commission. Neither the European Union nor the European Commission can
be held responsible for them.

## Citation

The users of this dataset are kindly asked to cite the following paper:

```bibtex
@misc{
  nousias2025malwaredetectiondockercontainers, 
  title={Malware Detection in Docker Containers: An Image is Worth a Thousand Logs},  
  author={Akis Nousias and Efklidis Katsaros and Evangelos Syrmos and Panagiotis Radoglou-Grammatikis and Thomas Lagkas and Vasileios Argyriou and Ioannis Moscholios and Evangelos Markakis and Sotirios Goudos and Panagiotis Sarigiannidis},
  year={2025}, 
  eprint={2504.03238},
  archivePrefix={arXiv}, 
  primaryClass={cs.CR}, 
  url={https://arxiv.org/abs/2504.03238}, 
}
```

## Contact

Contact [Panagiotis Radoglou-Grammatikis](mailto:[email protected]) for questions or comments.