Trust-based Authentication Aided Blockchain for Distributed Learning in UAV Swarms : Challenges and Solutions

The remarkable characteristics of unmanned aerial vehicles (UAVs), such as high flexibility and real-time data acquisition, make them a promising platform for providing services and applications like data collection and aerial computing. Furthermore, by integrating UAVs with decentralized learning, emerging machine learning (ML)-assisted services (e.g., urban sensing and natural resource exploration) can be enhanced. However, the limited computation and communication resources of UAVs and the potential security risks pose challenges to the reliability and security for distributed learning in UAV swarms. This necessitates suitable learning policy design and continuous supervision throughout the process of decentralized learning. In this article, motivated by the merits of blockchain, we propose a trust-based authentication aided blockchain for distributed learning in UAV swarms, along with a corresponding layered architecture to ensure effectiveness and security. This architecture includes a perception layer, an aerial layer, and a blockchain layer. Moreover, the trust values of UAVs are derived from their behaviors during the decentralized learning process. These values, combined with smart contracts, enable dynamic and robust authentication for UAV swarms. Based on this framework, we present its advantages and exemplified applications. Moreover, the potential challenges related to security and efficiency when applying the proposed framework in wireless networks are also discussed. Additionally, we present a case study on trust-based authentication aided blockchain for distributed learning in UAV swarms. This includes a detailed procedure and solutions to the associated challenges, demonstrating its performance in terms of learning efficiency and security. We finally discuss some open research directions regarding the blockchain-assisted decentralized learning for UAV swarms. © 2024 IEEE.