GEANT-Flow25: Network Flow Dataset for Network Traffic Analysis

Network traffic analysis, such as identifying and classifying traffic types, plays a crucial role in maintaining network service quality and network security. As traditional approaches face increasing challenges, the use of machine learning is gaining growing interest in the networking field. We present GEANT-Flow25, a network traffic flow dataset collected from IPFIX collectors within GÉANT’s pan-European research and education backbone network. The dataset captures flow records suitable for traffic classification and anomaly detection tasks. Labels have been added for specific virtual routing instances used by flows, making it possible to determine whether a flow is intended for the commodity Internet (IAS), Research and Education, or one of the two research projects LHCONE and COPERNICUS. The flow records are particularly suitable for complex supervised and unsupervised machine learning approaches, such as deep neural networks and graph neural networks.