Blockchain-Enabled Secure Federated Learning for Digital Twin Networks

With the development of digital twin (DT) technology, the DT network (DTN) has emerged to represent virtually the physical network by building physical entities as interactive DT models. However, the DTN faces the challenge of privacy leakage due to the collection of private data in the DT model building process. Additionally, massive data uploaded to the DTN for centralized processing leads to security issues of single-point failure and poisoning attacks. In this article, we propose block-chain-enabled secure federated learning (FL) for constructing a decentralized and privacy-preserving DTN. Specifically, we utilize the FL paradigm and homomorphic encryption method to locally build DT models at the ciphertext level. Then, we propose integrating blockchain into FL to construct a decentralized architecture for the DTN to avoid single-point failure. Moreover, we devise the proof-of-gradient consensus mechanism to facilitate trusted interactions among DT models to resist poisoning attacks. Finally, the security analysis and performance evaluations demonstrate the superiority of the proposed schemes.