August 25, 2024

Our Work “SemNFT: A Semantically Enhanced Decentralized Middleware for Digital Asset Immortality” Is Accepted By ACM MM’24

Recently, our work “SemNFT: A Semantically Enhanced Decentralized Middleware for Digital Asset Immortality” is accepted to be published in The 32nd ACM International Conference on Multimedia (MM’24).

Background

Non-Fungible Tokens (NFTs) have emerged as a pivotal digital asset, offering authenticated ownership of unique digital content. Despite it has gained remarkable traction, yet face pressing storage and verification challenges stemming from blockchain’s permanent data costs. Existing off-chain or centralized storage solutions, while being alternatives, also introduce notable security vulnerabilities.

Methodology

We present SemNFT, an innovative decentralized framework integrated with blockchain oracle middleware services, addressing these persistent NFT dilemmas.

The architecture of the protocol can be divided into two parts, the off-chain part and the on-chain part, as shown in Figure 1. In the off-chain, the primary focus involves two key tasks: training the autoencoder model and downcasting the float array. The autoencoder model training process plays a vital role in data compression and feature extraction, while the downcasting of the float array aims to convert floating-point numbers to integers for subsequent operations. On the other hand, the on-chain part is primarily dedicated to the minting of NFTs from the integer array, which are subsequently stored and managed on the blockchain. This process enables the unique identification and ownership tracking of individual NFTs within the decentralized ledger system.

Figure 7 illustrates the NFT ownership verification process of SemNFT in comparison to existing NFTs. Under the current NFT storage scheme, if we need to verify an NFT image with unknown ownership, we must seek the NFT owned by the user in the contract  through their wallet address. Subsequently, we retrieve the NFT resource’s metadata through the link recorded in the contract and then request the image link in the metadata to obtain and verify the NFT image. If a disconnection occurs during the link request process, the verification process will be interrupted. Even if users store their NFT image source data off-chain, it cannot be matched and verified with assets in the smart contract.

Evaluation

The reconstruction quality can be shown as Table 2, Figure 8, and Figure 9. The gas usage comparison shows in Figure 10. And the compression comparison shows in Table 3.

Conclusion

This paper presented SemNFT, a decentralized framework tackling pressing NFT storage and verification challenges. By leveraging decentralized middleware services equipped with autoencoders, the framework ensures not only the secure and perpetual storage of digital assets in the form of NFTs but also maintains their intrinsic qualities and representations, thereby safeguarding their essence throughout the pipeline. Our framework is also highly compatible with existing NFT standards, adapting well to the current NFT community and ecosystem. To verify its feasibility, we implemented the proposed framework and conducted an evaluation Evaluations evidenced feasibility and efficiency, attaining over 10:1 data compression ratio across diverse NFT datasets with high visual fidelity. SemNFT streamlines verification by reconstructing images on-demand, circumventing external dependencies. SemNFT emerges as a pioneering framework, intertwining the sophisticated capabilities of autoencoders with the immutable and decentralized characteristics of blockchain to forge a path towards digital asset immortality. Future research might delve deeper into optimizing the autoencoder algorithms, exploring additional use-cases for the framework, and evaluating the real-world impact and applications of ensuring digital asset immortality within various industries.

Early Access:

https://mypage.cuhk.edu.cn/academics/caiwei/paper/papers/LehaoLKSC2024.pdf

Cite this work:

Lehao Lin, Hong Kang, Xinyao Sun, and Wei Cai, “SemNFT: A Semantically Enhanced Decentralized Middleware for Digital Asset Immortality“, In 32nd ACM International Conference on Multimedia (MM’24), Melbourne, Australia, October 28-November 1, 2024.

@inproceedings{lin2024semnft,
  title={SemNFT: A Semantically Enhanced Decentralized Middleware for Digital Asset Immortality},
  author={Lin, Lehao and Kang, Hong and Sun, Xinyao and Cai, Wei},
  booktitle={Proceedings of the 32nd ACM International Conference on Multimedia},
  year={2024}
}