Digital Twin-Assisted Federated Learning With Blockchain in Multi-Tier Computing Systems

In Industry 4.0 systems, numerous resource-constrained Industrial Internet of Things (IIoT) devices frequently exchange data for model training, raising serious security and privacy concerns. To address these issues, this paper proposes a digital twin (DT) and blockchain-assisted federated learning (FL) framework. In the proposed scheme, fog devices with sufficient computational resources generate DTs for edge devices to support local model training. We formulate an FL delay minimization problem that considers both model transmission and synchronization delays while introducing cooperative jamming to secure DT synchronization. To solve this non-convex problem, a decomposition-based optimization algorithm is developed. By introducing upper bounds on local training delay and jamming effects as auxiliary variables, the problem is transformed into a convex form that can be efficiently decomposed and solved. Furthermore, a blockchain-based verification mechanism and a validator selection algorithm are designed to ensure model integrity and participant authentication during the FL process, thereby enhancing blockchain scalability. The final global model is obtained by aggregating verified local updates through deep learning techniques. Numerical results demonstrate that the proposed cooperative interference-based DT–blockchain-assisted FL framework outperforms benchmark schemes in execution time, block efficiency, and model accuracy.