Atom: Neural Traffic Compression with Spatio-Temporal Graph Neural Networks

Link to paper: [here]

P. Almasan, K. Rusek, S. Xiao, X. Shi, X. Cheng, A. Cabellos-Aparicio, P. Barlet-Ros

Contact: paulalmasan@gmail.com

Abstract

Storing network traffic data is key to efficient network management; however, it is becoming more challenging and costly due to the ever-increasing data transmission rates, traffic volumes, and connected devices. In this paper, we explore the use of neural architectures for network traffic compression. Specifically, we consider a network scenario with multiple measurement points in a network topology. Such measurements can be interpreted as multiple time series that exhibit spatial and temporal correlations induced by network topology, routing, or user behavior. We present Atom, a neural traffic compression method that leverages spatial and temporal correlations present in network traffic. Atom implements a customized spatio-temporal graph neural network design that effectively exploits both types of correlations simultaneously. The experimental results show that Atom can outperform GZIP's compression ratios by 50%--65% on three real-world networks.

Instructions to execute

See the execution instructions

Description

To know more details about the implementation used in the experiments contact: paulalmasan@gmail.com

Please cite the corresponding article if you use the code from this repository:

@inproceedings{10.1145/3630049.3630170, author = {Almasan, Paul and Rusek, Krzysztof and Xiao, Shihan and Shi, Xiang and Cheng, Xiangle and Cabellos-Aparicio, Albert and Barlet-Ros, Pere}, title = {Atom: Neural Traffic Compression with Spatio-Temporal Graph Neural Networks}, year = {2023}, isbn = {9798400704482}, publisher = {Association for Computing Machinery}, address = {New York, NY, USA}, url = {https://doi.org/10.1145/3630049.3630170}, doi = {10.1145/3630049.3630170}, abstract = {Storing network traffic data is key to efficient network management; however, it is becoming more challenging and costly due to the ever-increasing data transmission rates, traffic volumes, and connected devices. In this paper, we explore the use of neural architectures for network traffic compression. Specifically, we consider a network scenario with multiple measurement points in a network topology. Such measurements can be interpreted as multiple time series that exhibit spatial and temporal correlations induced by network topology, routing, or user behavior. We present Atom, a neural traffic compression method that leverages spatial and temporal correlations present in network traffic. Atom implements a customized spatio-temporal graph neural network design that effectively exploits both types of correlations simultaneously. The experimental results show that Atom can outperform GZIP's compression ratios by 50\%--65\% on three real-world networks.}, booktitle = {Proceedings of the 2nd on Graph Neural Networking Workshop 2023}, pages = {1–6}, numpages = {6}, keywords = {neural traffic compression, spatio-temporal graph neural networks}, location = {<conf-loc>, <city>Paris</city>, <country>France</country>, </conf-loc>}, series = {GNNet '23} }