Trusted Secret Sharing Scheme with Verifiable Computation Based on On-Chain and Off-Chain Collaboration for IoT

Blockchain provides an excellent solution to the centralized trust problems encountered in federated IoT data sharing platforms, such as data tampering and single-point attacks. However, its transparency also undermines the privacy of federated learning in model training and model sharing, thus creating the following challenges: (1) Federated model training needs to be conducted off-chain to ensure privacy. Since off-chain computing is not supervised by the blockchain, the trustworthiness of off-chain training models cannot be guaranteed. (2) Federated models shared on-chain expose intermediate parameters to the public. Malicious users can infer the original training data from these parameters, leading to privacy leakage at the model sharing level. To address these issues, this paper proposes a federal trusted secret sharing model based on on-chain and off-chain collaboration. In this model, we first design a federated trusted verification algorithm to ensure the trustworthiness of models trained off-chain. Then, based on the algorithm, combined with CKKS and threshold homomorphic encryption, a deep federated trusted secret sharing scheme based on on-chain and off-chain collaboration for IoT is constructed to ensure comprehensive trust and privacy in federated blockchain sharing schemes. Simulation experiments show that the scheme is feasible.