A Privacy-preserving Auction Mechanism for Learning Model as an NFT in Blockchain-Driven Metaverse

The Metaverse, envisioned as the next-generation Internet, will be constructed via twining a practical world in virtual form, wherein Meterverse service providers (MSPs) are required to collect massive data from users (MUs). In this regard, critical demand exists for MSPs motivate MUs contribute computing resources and while preserving user privacy. Federated learning (FL), privacy-preserving collaborative machine paradigm, can support distributed intensive computation Metaverse. work, we first investigate minting models into NFT with FL assistance (referred FL-NFT), such that stakeholders control ownership share economic value of user-generated content (UGC). Specifically, encouraged establish decentralized autonomous organization (i.e., MU-DAO) aggregate local mint FL-NFT. optimize strategies by formulating an imperfect information Stackelberg game (IISG) trade off cost benefit. We apply backward induction derive equilibrium solution. Then, construct multi-winner sealed-bid auction mechanism (PMS-AM), which Hidden Markov Model (HMM) assists choosing rational bidding according historical bids, double determines winners price Finally, numerical results based on theoretical analysis simulations demonstrate proposed PMS-AM increase quality FL-NFT achieve properties incentive mechanisms individual rationality compatibility.