| Class | Samples |
|---|---|
| Total | 5797 |
| Bit Torrent | 1217 |
| DNS | 1412 |
| VoIP | 1320 |
| IoT | 1848 |
Packet Vision
Network Traffic Classification
We propose a new public and free dataset containing network traffic images, specifically designed to evaluate and compare algorithms for image classification on-the-fly. PacketVision aimed to explore the Open Policy Interface (OPI), an interface that supports NASOR to establish network slicing across multiple administrative and technological domains.
The built dataset had processed through Packet Vision, an approach capable of creating images from raw-data of network packets, collected passively on network interfaces. Our method currently advances those found in state-of-the-art because we consider both header and payload to build pictures, avoiding pre-processing ones. This strategy aimed to provide security and privacy by transforming the raw-data packet into images while enabling opportunities to exploit image classification technics, especially CNNs.
Dataset
The dataset is structured as follows:
Download
The PacketVision dataset is available on Zenodo for academic research and non-commercial use. The dataset provides image representations generated from network packets and supports research on image-based network traffic classification, computer vision, deep learning, and intelligent network analysis.
-
Dataset access
The official PacketVision dataset release is available through Zenodo. Please use the Zenodo record to download the dataset, access metadata, and cite this resource in academic publications. -
Terms of use
PacketVision is released for academic research and non-commercial use. Users should cite the Zenodo record and the associated PacketVision publications when reporting results obtained with this dataset.
Citation
All documents and papers that report research using the PacketVision dataset must include an appropriate citation to the Zenodo record and to the associated PacketVision publications. See the related publications section .
Publications
- Moreira, R., Rodrigues, L. F., Rosa, P. F., Aguiar, R. L., Silva, F. O.
Enhancing dynamism in management and network slice establishment through deep learning
2021 International Conference on Information Networking (ICOIN). IEEE, 2021.
[BibTeX] - Moreira, R., Rodrigues, L. F., Rosa, P. F., Aguiar, R. L., Silva, F. O.
Packet Vision: a convolutional neural network approach for network traffic classification
33rd SIBGRAPI Conference on Graphics, Patterns and Images. IEEE, 2020.
[BibTeX] - Moreira, R., Rodrigues, L. F., Rosa, P. F., Silva, F. O.
Improving the network traffic classification using the Packet Vision approach
Workshop de Visão Computacional, 2020.
[BibTeX]
Contact
Please contact Rodrigo Moreira or Larissa Rodrigues with questions or comments.