Fast Attention Network approach for traffic classification on next-generation networks

The evolution of next-generation wireless networks and the growing diversity of data-intensive digital services are creating new challenges for traffic analysis and service-aware management in Beyond 5G and 6G networks. In this context, systematic analysis of network traffic is essential to identify service usage patterns and support efficient network operation. This paper introduces an advanced classification framework based on Fast Attention Networks to autonomously identify different service categories in Beyond 5G and 6G environments. By employing a high-performance attention mechanism that jointly refines channel, spatial, and frequency representations, the proposed model addresses the challenge of classification accuracy in massive data streams. Through a comparative assessment of various neural architectures, this study demonstrates the superior and more consistent efficacy of the proposed model in identifying specific service types, while also showing robust behavior under limited training data availability. In addition, the framework is conceived to support NWDAF-oriented analytics in B5G environments, enabling service-aware intelligence for network optimization. The results highlight a robust and scalable framework for service identification within the technical landscape of Beyond 5G and 6G networks.