A Two-Dimensional Hybrid Federated Learning Framework for Secure Data Cooperation of Multiple Network Service Providers

In the era of artificial intelligence of things (AIoT), the substantial user data collected by network service providers (NSPs) holds great potential for enhanced service provisioning. However, the data isolation between NSPs limits the full utilization of data. In this article, we investigate collaborative computing among NSPs to fully unlock the potential of heterogeneous cellular data, with comprehensive discussions on its advantages, applications, and challenges. We propose a two-dimensional hybrid federated learning (2DHFL) framework to facilitate collaborative computing among multiple NSPs, with ingenious integration of horizontal and vertical federated learning techniques. The 2DHFL framework can effectively address the challenges related to incomplete features and insufficient training data while maintaining data privacy. A case study on map matching task is presented with empirical data-driven experiment results to demonstrate the effectiveness of the 2DHFL framework compared with state-of-the-art benchmark schemes.