Ownership Tokenization and Incentive Design for Learning-based User-Generated Content

Qinnan Zhang, Zehui Xiong, Jianming Zhu, Sheng Gao, Wanting Yang, Dusit Niyato

Abstract: The Metaverse is established by twining a practical world in a virtual form, where users could become creators of learning-based user-generated content (UGC). However, the digital assets are not owned by the creators, which deviates from the decentralized Metaverse based on blockchain in Web 3.0. The contribution incentive design for UGC still faces the challenges of ownership verification and privacy protection. To this end, we propose minting the learning model into the non-fungible token (NFT) with federated learning (FL) assistance (referred to as FL-NFT), such that users as stakeholders can control the ownership and share the economic value of UGC. Specifically, the users are encouraged to establish a decentralized autonomous organization (DAO) to aggregate local models and mint FL-NFT. We formulate an auction interaction of FL-NFT as imperfect information Stackelberg game (IISG) to optimize the bidding strategies to realize individual rationality. Finally, we conduct simulations to show the effectiveness of the proposed scheme.