Federated Learning Framework with Straggling Mitigation and Privacy-Awareness for AI-based Mobile Application Services

This work proposes a novel framework to address straggling and privacy issues for federated learning (FL)-based mobile application services, considering limited computing/communications resources at users (MUs)/mobile provider (MAP), cost, the rationality incentive competition among MUs in contributing data MAP. Particularly, MAP first determines set of best FL process based on MUs' provided information/features. Then, each selected MU can encrypt part local upload encrypted an training process, addition process. For that, propose contract according its expected data. To find optimal contracts that maximize utilities while maintaining high quality system, we develop multi-principal one-agent contract-based problem MAP's computing resources, asymmetric information between MUs. Experiments with real-world dataset show our speed up time 49% improve prediction accuracy 4.6 times enhancing network's social welfare 114% under cost consideration compared those baseline methods.