Recently, our paper titled “Exploring the application of Time Series Foundation Models to network monitoring tasks” co-authored by our chair in collaboration with researchers from the University of Würzburg and the Austrian Institute of Technology (AIT), has been accepted to the 269th Volume of the Computer Networks journal. This paper investigates how Time Series Foundation Models (TSFMs), can revolutionize network monitoring. Traditional machine learning approaches often require extensive manual setup and struggle to adapt to new scenarios. In contrast, TSFMs are pre-trained on diverse time-series data and therefore offer flexible, plug-and-play solutions that work well even with minimal task-specific training.
These models are tested on a real-world challenge to estimate the video streaming Quality of Experience (QoE) from encrypted traffic. The findings show that TSFMs can deliver strong performance without prior task-specific training and improve further with just a few example cases. This research highlights the potential of TSFMs to simplify and scale AI-driven network monitoring across a range of use cases.

Paper: Nikolas Wehner, Pedro Casas, Katharina Dietz, Stefan Geißler, Tobias Hoßfeld, Michael Seufert. "Exploring the application of Time Series Foundation Models to network monitoring tasks." 269th Computer Networks Journal.

Link to paper:
Exploring the application of Time Series Foundation Models to network monitoring tasks