Industrial blockchain threshold signatures in federated learning for unified space-air-ground-sea model training

作者：Jingxue Chen, Eric Wang, Gautam Srivastava, Turki Ali Alghamdi, Fazlullah Khan, Saru Kumari, Hu Xiong · 发表于：Journal of Industrial Information Integration · 年份：2024 · DOI：10.1016/j.jii.2024.100593 · 被引用次数：54 · 研究领域：Privacy-Preserving Technologies in Data、Cryptography and Data Security、Blockchain Technology Applications and Security

The space-air-ground-sea three-dimensional (3D) network is a comprehensive communication network system. This 3D network combines the extensive coverage of satellite communications, the adaptability of unmanned aerial vehicle (UAV) communications, the reliability of terrestrial communications, and the necessity of maritime communications. These networks generate enormous amounts of data, and training machine learning (ML) models on this data will have a significant impact on the industry. At the same time, the availability of such data poses numerous security threats, which can be overcome by Federated Learning (FL). The decentralized training in FL can provide a universal model from local data generated by the 3D network However, most existing FL frameworks have a centralized server, which questions the credibility, single-point failure, and global confidence. To solve these problems, industrial blockchain technology has received much attention by replacing centralized servers in traditional FL, which offers a promising approach to address key issues such as data privacy and security. In a blockchain-based system, digital signature is the core component for ensuring data integrity and system security, however, private key disclosure can pose significant risks. The security can be enhanced by using threshold signature, which provides a more reliable foundation for FL by storing keys in multiple nodes and requiring multiple nodes to collaborate to generate signatures. In this ...