BFKD: Blockchain-Based Federated Knowledge Distillation for Aviation Internet of Things

【Author】 Deng, Wu; Li, Xinyan; Xu, Junjie; Li, Weihan; Zhu, Guangtian; Zhao, Huimin

【Source】IEEE TRANSACTIONS ON RELIABILITY

【影响因子】5.883

【Abstract】Aviation Internet of Things (AIoT) data sharing can create tremendous value for participants. With the development of AIoT and intelligent civil aviation, data security and privacy protection have become more prominent. To improve the data security and privacy of AIoT, and allow heterogeneous devices to participate in federated training, in this article, a blockchain-based federated knowledge distillation (BFKD) is proposed. First, federated knowledge distillation is used for knowledge transferring between clients to protect client data privacy. Second, a clustering algorithm is used to select aggregated soft labels to enhance Byzantine resistance. Finally, the consortium blockchain validates the data exchange process of federated knowledge distillation. In addition, a federated score grouping practical Byzantine fault-tolerant algorithm is proposed for improving consensus efficiency and to encourage participants to contribute more public data and honestly participate in federated training. Theoretical analysis and experimental results show that BFKD boasts high data mining performance and communication efficiency while safeguarding data privacy and security.

【Keywords】Blockchains; Data privacy; Internet of Things; Data models; Training; Servers; Resistance; Data security; Accuracy; Resilience; Aviation Internet of Things (AIoT); blockchain; federated knowledge distillation (FKD); federated score grouping practical Byzantine fault-tolerant (FSGPBFT); incentive mechanism

【发表时间】2024 2024 OCT 24

【收录时间】2024-11-09

【文献类型】案例研究

【主题类别】

区块链应用-实体经济-航空航天