Blockchain-Based Decentralized Learning for Security in Digital Twins

This work aims to analyze malicious communication behaviors that pose a threat to the security of digital twins (DTs) and safeguard user privacy. A unified and integrated multidimensional DTs Network (DTN) architecture is constructed. On this basis, the propagation process model of malware in the network is built to analyze the malicious propagation behavior that threatens network security. This model ensures the protection of mobile distributed machine learning system security. Blockchain technology is a distributed data protection mechanism with broad prospects. It is characterized by decentralization, transparency, and anonymity, which can help ensure secure network data sharing and privacy protection. Based on this, this work designs a secure distributed data sharing (DDS) architecture based on blockchain to improve the security and reliability of data protection with the support of the Internet of Things (IoT). Then, digital resource allocation based on semi-distributed learning is examined to propose a broad learning federated continuous learning (BL-FCL) algorithm combining blockchain and DTs. This algorithm significantly speeds up the model training process. Broad learning technology supports incremental learning. In this way, each client does not need to retrain when learning the newly generated data. In the experimental part, the prediction accuracy of BL-FCL on the mixed national institute of standards and technology data set is similar to that of the FedAvg-50 and FedAvg-80 schemes. As the number of devices increases from 1 to 6, the detection probability exhibits a rapid decrease. However, as the number of devices further increases from 6 to 10, the detection probability gradually decreases at a slower rate until it reaches 0. Comparatively, the prediction accuracy of the BL-FCL outperforms the federated averaging algorithm-based scheme by 20%-60%. The BL-FCL reported here can deal with the problem of inaccurate training while ensuring the privacy and security of users. This work is of great significance for ensuring the security of the DTN and promoting the development of the digital economy. The results can provide references for applying blockchain and distributed learning in the DT field.