Toward a Secure Edge-Enabled and Artificially Intelligent Internet of Flying Things Using Blockchain

Unmanned aerial vehicles (UAVs) with artificial intelligence (AI) have ushered in a new era of mobile edge computing (MEC). Traditional AI models based on the aggregation of UAV sensing data that mostly contain private and sensitive user data, may raise serious privacy and data misuse problems. Federated learning (FL), as a potential distributed AI paradigm, has allowed UAVs to jointly train a global model without exposing their local sensing data. In the general context, flying things communicate only local model updates. The training process ends when the global model reaches a certain threshold. The centralized curator aggregation model can be subject to diverse threats such as DDoS and single point of failure. To improve the security of FL implementation and create a privacy-preserving model, we consider a new blockchain empowered AI paradigm using both wireless miners and edge computing at flying things for security requiring heterogeneous vehicular systems. In our proposal, blockchain is the backbone allowing secure and safe data exchange between them, providing decentralized FL training without the need for a central server. The proposed architecture is apt to cope with FL privacy leakage, insider/outsider attacks, malicious opponents, inference, and poisoning. It transforms intelligent MEC systems into decentralized, secure, and privacy enhanced networks.