A Blockchain-Enabled AIoT Framework for Secure Metaverse in Wireless Communication Networks

The wireless communication network, comprising billions of cloud, edge, and end devices, enables emerging applications such as the metaverse—an immersive virtual environment where experiences and user interactions converge. However, user interactions in the metaverse generate substantial amounts of private data (e.g., identity, location), raising significant privacy concerns despite their utility in training machine learning models. Artificial Intelligence of Things (AIoT) offers solutions to these challenges, with Federated Learning (FL) serving as a decentralized framework that enables collaborative model training without exposing local data. Nevertheless, deploying FL in large-scale environments like the metaverse increases vulnerability to malicious attacks. To address this issue, we propose a blockchain-based FL architecture that enhances trust and security. The architecture integrates a multi-task FL strategy with blockchain sharding to boost system throughput and reduce resource consumption. By partitioning the blockchain into smaller shards, we lower computational demands and enable concurrent training of multiple models, improving efficiency. We also design a shard creation algorithm based on bipartite matching and a bandwidth scheduling mechanism that prioritizes reliable devices with informative data. Experimental results show that our architecture outperforms existing baselines across multiple evaluation metrics.