Multi-party credible AI model transaction settlement method and system based on block chain technology

Abstract

The invention provides a multi-party credible AI model transaction settlement method and system based on a block chain technology, and belongs to the technical field of block chain intelligent contract transaction management, and the method comprises the steps: S1, uploading a model mirror image to a far-end mirror image warehouse, generating a unique URI, and storing an uploading record on a block chain; s2, after the model provider receives the transaction request, casting an NFT representative model use right through the smart contract, and issuing the NFT representative model use right to the model provider; s3, the model purchaser pays ERC20 tokens to the model provider, and the model provider transfers NFT to the model purchaser; and S4, using asymmetric encryption and an ERC721 smart contract to verify the ownership of the model purchaser to the wallet address and the wallet address has a corresponding NFT, and if verification is passed, creating a temporary account and issuing the temporary account to the model purchaser. According to the method, the security and transparency in the model transaction process are enhanced, and the credibility of a model provider and a user to the platform is improved.

Description

Technical Field

[0001] This invention belongs to the field of trading and management of artificial intelligence models, and specifically relates to a multi-party trusted AI model trading settlement method and system based on blockchain technology. Background Technology

[0002] With the development of artificial intelligence (AI) technology, more and more enterprises and individual developers need access to high-quality pre-trained models to accelerate their project development process. However, existing AI model distribution methods typically rely on centralized platforms and services, which presents several challenges:

[0003] Security issues: Under traditional distribution models, it is difficult to guarantee the security and integrity of the model, making it vulnerable to unauthorized access or tampering. For example, once a mirror file is made public, anyone can directly download and use the model via its URI without any verification or authorization process. This openness makes the model extremely susceptible to malicious attacks and abuse, thereby affecting its integrity and security.

[0004] Copyright protection and the issue of digital asset replicability: Current centralized platforms lack effective means to track and verify model ownership and usage rights, leading to difficulties in intellectual property protection. Model images, as a type of digital asset, are highly replicable (a single image can be pulled countless times), making it opaque to the provider how many copies of their uploaded digital assets have been copied and sold. Furthermore, providers cannot know exactly how many users have actually used their models, nor can they ensure that each use has been compensated accordingly. If platforms conceal the number of transactions for profit, it will severely infringe upon the interests of providers, leading to a decrease in their trust in the platform.

[0005] The issue of unclear model usage rights: Existing centralized platforms have significant shortcomings in managing AI model usage rights. Centralized service-based usage rights management often relies on traditional database storage to record, track, and query usage rights, which is vulnerable to external attacks or internal data tampering. These platforms typically lack detailed usage rights recording mechanisms, failing to accurately track the specific users and scope of use for each model, resulting in unclear usage rights records. Simultaneously, the lack of transparency in the transfer of model usage rights makes it prone to disputes and conflicts. Furthermore, platforms struggle to flexibly adjust user permissions, such as setting expiration dates or restricting specific uses, further exacerbating the chaos in usage rights management. Due to the lack of transparency in usage rights recording and management, there is a lack of sufficient trust between providers and purchasers, severely impacting the healthy development of the entire ecosystem. These problems collectively lead to unclear model usage rights, harming the interests of all parties and hindering the widespread application of the platform.

[0006] Therefore, a decentralized, highly reliable, and transparent AI model trading mechanism is needed to improve the credibility and security of model trading platforms. The application of blockchain technology can ensure that every transaction is accurately recorded and tamper-proof, providing clear records of usage rights and a transparent transfer process, enhancing platform trust, and offering users a more convenient and secure trading environment. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides a multi-party trusted AI model transaction settlement method based on blockchain technology, comprising the following steps:

[0008] Step S1: Upload the model image to the remote image repository, generate a unique URI, and store the upload record on the blockchain;

[0009] Step S2: After receiving the transaction request, the model provider uses a smart contract to mint an NFT representing the right to use the model according to the ERC721 standard, and issues it to the model provider.

[0010] Step S3: The model purchaser pays the model provider with ERC20 tokens, and the model provider transfers the NFT to the model purchaser to represent the right to use the model image sold.

[0011] Step S4: Using asymmetric encryption and ERC721 smart contracts, verify the model purchaser's ownership of the wallet address and the corresponding NFT. If the verification is successful, create a temporary account and issue it to the model purchaser.

[0012] Beneficial effects:

[0013] 1. This invention is the first to combine non-fungible tokens (NFTs) with the right to use AI models, enabling the creation, tracking, and verification of the right to use AI models, thus solving the problems of AI model usage rights being easily tampered with and difficult to create, track, and verify.

[0014] 2. This invention proposes an AI model usage right verification mechanism based on asymmetric encryption and smart contracts, which realizes the process of obtaining AI model image retrieval permission by virtue of the user's NFT model usage right, and solves the problem of interaction authentication between centralized image management system and blockchain smart contract.

[0015] 3. This invention uses ERC20 tokens as the circulating tokens of a multi-party AI model trading platform, which solves the problem that transaction information may be concealed by the platform due to the fact that AI models, as digital assets, can be copied and sold multiple times. This ensures that every transaction record can be publicly and transparently queried in the blockchain and is tamper-proof. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a multi-party trusted AI model transaction settlement method based on blockchain technology according to the present invention;

[0017] Figure 2 This is a schematic diagram of the architecture of a multi-party trusted AI model transaction settlement method based on blockchain technology according to the present invention;

[0018] Figure 3 This is a structural block diagram of a multi-party trusted AI model transaction settlement system based on blockchain technology according to the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0020] Example 1

[0021] like Figure 1 As shown in the figure, an embodiment of the present invention provides a multi-party trusted AI model transaction settlement method based on blockchain technology, which includes the following steps:

[0022] Step S1: Upload the model image to the remote image repository, generate a unique URI, and store the upload record on the blockchain;

[0023] Step S2: After receiving the transaction request, the model provider uses a smart contract to mint an NFT representing the right to use the model according to the ERC721 standard, and issues it to the model provider.

[0024] Step S3: The model purchaser pays the model provider with ERC20 tokens, and the model provider transfers NFTs to the model purchaser to represent the right to use the model image sold.

[0025] Step S4: Using asymmetric encryption and ERC721 smart contracts, verify the model purchaser's ownership of the wallet address and the corresponding NFT. If the verification is successful, create a temporary account and issue it to the model purchaser.

[0026] In one embodiment, step S1 above—uploading the model image to a remote image repository, generating a unique URI, and storing the upload record on the blockchain—specifically includes:

[0027] Step S11: The model provider uses a command to package the model container image to be uploaded into a tar archive image file;

[0028] The model provider uses the `docker save` command in their local container management environment to package the AI ​​model container image to be uploaded into a `.tar` file. For example, execute the following command:

[0029] docker save -o<model_name> .tar<container_name>

[0030] This .tar file contains all the model's dependencies and environment configurations, ensuring that the image can run seamlessly in other environments.

[0031] Implementation details:

[0032] Model providers need to ensure that Docker is installed on their local environment and that the model container to be uploaded has been built.

[0033] After packaging, the model provider uploads the .tar file through the system's front-end interface, and the system back-end receives and processes the file.

[0034] Step S12: Load the tar package image file and tag it; create a separate project for each model image to isolate permissions for different images, push the image file to the image repository, and generate a unique URI;

[0035] After receiving the .tar file, the system loads the image using the `docker load` command and tags it. Then, the system uses Harbor's RESTful API to create a separate project for each model image, with the project name generated using a UUID to ensure uniqueness. Finally, the system uses the `docker push` command to push the image to the Harbor repository, generating a unique image URI.

[0036] Implementation details:

[0037] System call docker load <<model_name> Load the image using .tar and use docker tag<image_id> harbor.example.com / <project_name> / <model_name> : <tag>Label it.

[0038] Harbor projects are created by calling their RESTful API, for example by sending a POST request to https: / / harbor.example.com / api / v2.0 / projects, specifying the project name and permission settings.

[0039] Once the image is pushed out, the system generates a unique URI, such as harbor.example.com / 7ft3lw / yolov5:latest, to identify the location of the image.

[0040] Step S13: The model provider provides the model pricing, which is in tokens per day, uploads the model information to the blockchain, and generates a model ID; the model information includes: uploader's wallet address, upload time, image URI, and image pricing.

[0041] The model provider sets the model price (in AMT / day) and records key model information on the blockchain. This information includes, but is not limited to, the uploader's wallet address, upload time, image URI, and image price.

[0042] Implementation details:

[0043] The system calls the recordModelUpload(address uploader, string memoryuri, uint256 pricePerDay) function in the smart contract to write the relevant information to the blockchain.

[0044] The pricing unit is represented by ERC20 standard tokens (AMT), and the smart contract is implemented using the @openzeppelin / contracts / token / ERC20 / ERC20.sol library.

[0045] Records on the blockchain are immutable and can be publicly viewed by all transaction participants.

[0046] In one embodiment, step S2 above, after receiving the transaction request, involves the model provider using a smart contract to mint an NFT representing the right to use the model according to the ERC721 standard, and then issuing it to the model provider. This specifically includes:

[0047] Step S21: Generate an NFT metadata file based on the transaction request. The metadata includes: image URI and expiration time of the right of use.

[0048] The system generates an NFT metadata file based on the model's relevant information. The metadata file is in JSON format and contains key information such as the image URI and expiration date of the usage rights. For example: { "uri": "harbor.example.com / 7ft3lw / yolov5:latest", "expired_at": 1733068800 }

[0049] Implementation details:

[0050] The system reads the model's upload records (such as image URIs and pricing information) from the blockchain and calculates the expiration time of the usage rights based on the purchase duration.

[0051] The content of the metadata file is dynamically generated by the backend, ensuring that the NFT metadata corresponding to each transaction is unique.

[0052] Step S22: Upload the metadata file to the IPFS interplanetary file system to obtain the file's CID;

[0053] The system uploads the generated NFT metadata file to IPFS (InterPlanetary File System) and obtains the file's CID (Content Identifier). For example, using Kubo as the IPFS implementation, the following command is executed:

[0054] ipfs add metadata.json

[0055] After a successful upload, the system will obtain a unique CID, in the form of:

[0056] QmPK1s3pNYLi9ERiq3BDxKa4XosgWwFRQUydHUtz4YgpqB.

[0057] Implementation details:

[0058] IPFS nodes are deployed on the system backend to ensure the stability of file uploads and access.

[0059] As a unique identifier for files in IPFS, CID features decentralized storage and immutability.

[0060] Step S23: Based on the metadata, mint an ERC721 standard non-fungible token (NFT) representing the right to use the model on the blockchain. The tokenURI of the NFT is set to CID and points to the metadata file in IPFS.

[0061] The system invokes an ERC721 standard smart contract to mint an NFT on the blockchain representing the right to use the model. The tokenURI field of this NFT points to the CID metadata file in IPFS.

[0062] Implementation details:

[0063] Use of smart contracts.

[0064] The library @openzeppelin / contracts / token / ERC721 / extensions / ERC721URIStorage.sol implements a custom awardItem function: function awardItem(address player, string memory tokenURI) publicreturns (uint256) { uint256 newItemId = _tokenIdCounter.current(); _tokenIdCounter.increment(); _mint(player, newItemId); _setTokenURI(newItemId, tokenURI); return newItemId; }

[0065] In this embodiment of the invention, the player parameter points to the wallet address of the model provider, and the tokenURI parameter points to the CID of the metadata file in IPFS.

[0066] Once minted, the ownership of the NFT belongs to the model provider, and detailed information about the NFT can be found through a blockchain explorer.

[0067] In one embodiment, step S3 above—where the model purchaser pays the model provider with ERC20 tokens and the model provider transfers NFTs to the model purchaser to represent the right to use the sold model image—specifically includes:

[0068] Step S31: The model buyer initiates a transaction and provides transaction information, including the model ID and purchase duration, into the smart contract;

[0069] The model buyer selects an AI model of interest through a front-end interface and enters transaction information such as the purchase duration (e.g., number of days). This information is encapsulated into a transaction request and sent to the smart contract. Specifically, the front-end application calls the smart contract's `purchase(uint256 modelId, uint256 duration)` function, where `modelId` is the unique identifier of the selected model, and `duration` represents the purchase duration.

[0070] Implementation details:

[0071] The model purchaser first logs into the system and connects their Ethereum wallet (such as MetaMask) to ensure they have enough ERC20 tokens (AMT) for payment.

[0072] After the user selects a model, the front-end application generates a transaction request object containing the model ID and duration based on the user's selection, and sends it to the smart contract on the blockchain through the Web3.js or Ethers.js library.

[0073] Step S32: Calculate the transaction order amount based on the purchase duration, unit price, and platform commission, and initiate an ERC20 token transfer authorization application to the model purchaser;

[0074] Once a transaction request is received, the smart contract automatically calculates the total transaction amount based on factors such as the model's pricing, purchase duration, and potential platform commission. Next, the smart contract initiates an ERC20 token transfer authorization request to the model purchaser, requiring the purchaser to confirm and approve the transfer of the corresponding number of AMT tokens from their wallet to the model provider.

[0075] Implementation details:

[0076] The smart contract contains a billing logic that queries the price of the model (in AMT / day) based on the model ID, and then multiplies it by the purchase duration to calculate the total amount.

[0077] After the calculation is complete, the smart contract will call the approve(address spender, uint256 amount) method in the ERC20 token contract to request the buyer to authorize the smart contract to transfer the specified number of AMT tokens on their behalf.

[0078] The front end will prompt the user to confirm this authorization operation. Usually, this step requires the user to manually click to confirm in MetaMask or other wallet plugins.

[0079] Step S33: The smart contract transfers the ERC20 tokens of the model buyer to the model provider, completing the payment process of the transaction;

[0080] After obtaining authorization from the buyer, the smart contract will execute the actual transfer of ERC20 tokens, transferring a specified number of AMT tokens from the buyer's account to the model provider's account. This process is accomplished by calling the `transferFrom(address sender, address recipient, uint256 amount)` method of the ERC20 token contract.

[0081] Implementation details:

[0082] The smart contract acts as a "spender," using the authorization previously granted by the buyer to directly deduct the corresponding amount of AMT tokens from the buyer's account and transfer them to the model provider's account.

[0083] Once a transaction is successful, the smart contract records the transfer on the blockchain, ensuring that all operations are open, transparent, and tamper-proof.

[0084] Step S34: Transfer the NFT from the model provider to the model purchaser to complete the model usage rights transaction.

[0085] The final step is to transfer the NFT, representing the right to use the model, from the model provider to the model buyer. The smart contract calls the `safeTransferFrom(address from, address to, uint256tokenId)` method in the ERC721 standard to transfer ownership of the NFT from the model provider to the buyer, thus officially completing the entire transaction process.

[0086] Implementation details:

[0087] The smart contract obtains the tokenId of the NFT and securely transfers the NFT from the model provider's wallet address to the buyer's wallet address using the safeTransferFrom method.

[0088] After a successful transfer, the buyer can view the newly acquired NFT through their Ethereum wallet, proving their right to use the specific AI model image.

[0089] At the same time, the system will create a temporary Robot account with pull permissions for the buyer based on the metadata file in the NFT (stored in IPFS). The expiration time of this account is the same as the time when the model buyer purchased it, allowing them to access the model image in the Harbor repository.

[0090] In one embodiment, step S4 above: using asymmetric encryption and an ERC721 smart contract, verifying the model purchaser's ownership of the wallet address and that the wallet address owns the corresponding NFT. If the verification passes, a temporary account is created and issued to the model purchaser, specifically including:

[0091] Step S41: The model purchaser initiates a request for model usage credentials, which includes their wallet address in the blockchain and the NFT's Token ID;

[0092] The model purchaser sends a request to the model image management module to obtain a certificate of use for the model. This request includes the wallet address on the blockchain and the token ID of the NFT held.

[0093] Implementation details:

[0094] The model purchaser submits a request through the front-end interface, and the front-end application calls the back-end API interface, such as POST / api / validate-model-access.

[0095] The request body includes: { "wallet_address": "0xYourWalletAddress", "token_id": 12345 }

[0096] After receiving the request, the backend passes this information to the subsequent signature verification and NFT ownership verification logic.

[0097] Step S42: The model purchaser signs the request using their private key and publishes their public key to the blockchain using their wallet account for signature verification;

[0098] The model purchaser uses their private key to sign the request, generating a digital signature to prove their ownership of the wallet address. The purchaser then publishes their public key to the blockchain for the system to verify the authenticity of the signature.

[0099] Implementation details:

[0100] Use the Web3.js or Ethers.js library to generate the signature. For example:

[0101] const message = "Validate ownership of wallet address and NFT";

[0102] const signature = await web3.eth.personal.sign(message,walletAddress);

[0103] After the signature is generated, the model purchaser publishes its public key to the blockchain via MetaMask or other wallet plugins (e.g., through smart contract storage).

[0104] The purpose of public key distribution is to ensure that the system can verify the authenticity and legitimacy of signatures.

[0105] Step S43: Verify the signature of the request. If the signature verification is successful, verify with the smart contract whether the wallet address owns the corresponding NFT.

[0106] First, the signature provided by the model purchaser is verified to confirm their ownership of the wallet address. Upon successful signature verification, the smart contract is queried to determine if the wallet address holds the corresponding NFT.

[0107] Implementation details:

[0108] The signature verification process is implemented using public keys and signatures on the blockchain. For example: function verifySignature(address signer, bytes32 messageHash, bytesmemory signature) public pure returns (bool) { return signer == ECDSA.recover(messageHash, signature); }

[0109] After successful signature verification, the ownerOf(uint256 tokenId) method of the ERC721 smart contract is called to verify whether the wallet address is the holder of the NFT with the specified token ID.

[0110] If verification fails, an error message will be returned, indicating that the user does not have permission to access the model.

[0111] Step S44: If the verification is successful, create a temporary account with only pull permissions for the project containing the model, and return it to the model purchaser as a credential for using the model.

[0112] Once signature verification and NFT ownership check are successful, the corresponding model image project in the Harbor repository is located based on the URI information in the NFT metadata. A temporary Robot account with only pull permissions is created for this project, and this account information is returned to the model purchaser as proof of model usage.

[0113] Implementation details:

[0114] Call Harbor's RESTful API to create a Robot account with pull permissions for the project containing the model. For example:

[0115] POST https: / / harbor.example.com / api / v2.0 / robots

[0116] The request body includes: { "name": str(uuid.uuid4())[:8], "description": "Project puller. Created for imagebuyer.", "level": "project", "duration": -1, "permissions": [ { "access": [ { "resource": "repository", "action": "pull", "effect": None } ], "kind": "project", "namespace": project_name, } ] }

[0117] After successful creation, Harbor returns the name and key of the Robot account, for example: { "name": "robot$project+tempuser", "secret": "iAhgxyKyP3D4rlFJh7AfyKhrBgjPKtpT" }

[0118] Figure 2 A schematic diagram of the architecture of the method of the present invention is shown.

[0119] Example 2

[0120] like Figure 3 As shown, this embodiment of the invention provides a multi-party trusted AI model transaction settlement system based on blockchain technology, including the following modules:

[0121] The model image management module 51 is used to process and store model image files, verify users' right to use the model, and create and manage image usage credentials.

[0122] This module is used to process and store model image files, including: managing uploaded AI model container images to ensure their secure storage and efficient access; verifying users' rights to use the models, including: using asymmetric encryption technology and ERC721 smart contracts to verify whether users own the corresponding NFTs; and creating and managing image usage credentials, including: providing users with temporary Robot accounts with pull permissions based on the verification results.

[0123] The model image management module includes the following modules:

[0124] The backend business view unit is used to process business requests sent from the frontend, call functions of other units in this module, and perform file management and database operations; for example, it uses the Django framework to build API interfaces, receive frontend requests, and call the corresponding business logic processing functions.

[0125] The container operation unit is used to load uploaded container image files, add tags, and push images to the image repository; for example, it uses `docker load` to load the image, `docker tag` to add tags, and uses the Harbor RESTful API to push the image to the repository.

[0126] The image repository unit is used to store container images, isolate permissions for different images, and create temporary accounts for image pull operations; for example, the Harbor API can be used to create an independent project for each model image, set permissions, and generate a unique Robot account.

[0127] The smart contract transaction module 52 is used to process transaction requests, perform balance inquiries, authorizations, and transfers of ERC20 tokens, mint, transfer, and verify NFTs, and store transfer records on the blockchain;

[0128] The smart contract transaction module includes the following modules:

[0129] ERC20 Circulating Token Management Contract: Used to issue, query, and transfer ERC20 standard tokens circulating within the system, and to authorize the main business logic contract to operate on the tokens; for example, using the ERC20.sol contract template in the OpenZeppelin library to define the basic functions and authorization mechanism of the tokens.

[0130] ERC721 Nonfungible Token Management Contract: Used to mint, query, transfer, and verify nonfungible tokens representing the right to use a model within the system, and to authorize operable NFT tokens to the main business logic contract; for example, using the ERC721URIStorage.sol contract template in the OpenZeppelin library to define the minting, querying, and transfer functions of NFTs.

[0131] IPFS (InterPlanetary File System) Units: Used to store NFT metadata files and provide a CID as the NFT's URI pointing to the target; for example, using Kubo as an IPFS node, upload a JSON-formatted metadata file through the IPFS API to obtain a unique CID.

[0132] Transaction business logic smart contracts: used to handle the main transaction requests, transfer authorized ERC20 tokens and NFTs, and complete the transaction; for example, defining smart contract functions such as initiatePurchase, transferToken, etc., to handle the entire process from initiating a transaction to completing payment and NFT transfer.

[0133] Blockchain Explorer Unit: Used to query token transfer records within the smart contract transaction module, making transaction information publicly and transparently displayed to any user; for example, integrating Blockscout or other blockchain explorers to allow users to view detailed information for each transaction.

[0134] The front-end user interface module 53 is used to provide users with a user-friendly interface, allowing them to log in to their Ethereum wallets and interact directly with smart contracts.

[0135] The front-end user interface module includes the following modules:

[0136] View control unit: Responsible for displaying basic page elements and interacting with the model mirror management module; for example, using the Vue.js framework to build front-end pages and combining UI libraries such as Prime Vue to improve user experience.

[0137] Ethereum Wallet Management Unit: Used to support logging into Ethereum wallets, interacting with smart contracts using the Application Binary Interface (ABI), and providing a blockchain-related operation interface; for example, integrating the MetaMask plugin, interacting with smart contracts through the Web3.js or Ethers.js library, and realizing functions such as wallet connection, signing, and transactions.

[0138] A blockchain-based multi-party trusted AI model transaction settlement device includes one or more electronic devices, wherein the one or more electronic devices are used to implement a blockchain-based multi-party trusted AI model transaction settlement method, system, and device.

[0139] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.< / tag>

Claims

1. A multi-party trusted AI model transaction settlement method based on blockchain technology, characterized in that, include: Step S1: Upload the model image to the remote image repository, generate a unique URI, and store the upload record on the blockchain; Step S2: After receiving the transaction request, the model provider uses a smart contract to mint an NFT representing the right to use the model according to the ERC721 standard, and issues it to the model provider. Step S3: The model purchaser pays the model provider with ERC20 tokens, and the model provider transfers the NFT to the model purchaser to represent the right to use the model image sold. Step S4: Using asymmetric encryption and ERC721 smart contracts, verify the model purchaser's ownership of the wallet address and the corresponding NFT. If the verification is successful, create a temporary account and issue it to the model purchaser.

2. The multi-party trusted AI model transaction settlement method based on blockchain technology according to claim 1, characterized in that, Step S1: Upload the model image to a remote image repository, generate a unique URI, and store the upload record on the blockchain. Specifically, this includes: Step S11: The model provider uses a command to package the model container image to be uploaded into a tar archive image file; Step S12: Load the tar package image file and tag it; create a separate project for each model image to isolate permissions for different images, push the image file to the image repository, and generate a unique URI; Step S13: The model provider provides the model pricing, which is in tokens per day, uploads the model information to the blockchain, and generates a model ID; the model information includes: uploader's wallet address, upload time, image URI, and image pricing.

3. The multi-party trusted AI model transaction settlement method based on blockchain technology according to claim 2, characterized in that, Step S2: After receiving the transaction request, the model provider uses a smart contract to mint an NFT representing the right to use the model according to the ERC721 standard, and issues it to the model provider. Specifically, this includes: Step S21: Generate an NFT metadata file based on the transaction request. The metadata includes: image URI and expiration time of the right of use. Step S22: Upload the metadata file to the IPFS interplanetary file system to obtain the file's CID; Step S23: Based on the metadata, mint an ERC721 standard non-fungible token (NFT) representing the right to use the model on the blockchain. The tokenURI of the NFT is set to the CID and points to the metadata file in IPFS.

4. The multi-party trusted AI model transaction settlement method based on blockchain technology according to claim 3, characterized in that, Step S3: The model purchaser pays the model provider with ERC20 tokens, and the model provider transfers the NFTs to the model purchaser to represent the right to use the sold model image. Specifically, this includes: Step S31: The model buyer initiates a transaction and provides transaction information, including the model ID and purchase duration, into the smart contract; Step S32: Calculate the transaction order amount based on the purchase duration, unit price, and platform commission, and initiate an ERC20 token transfer authorization application to the model purchaser; Step S33: The smart contract transfers the ERC20 tokens from the model purchaser to the model provider, completing the payment process for the transaction; Step S34: Transfer the NFT from the model provider to the model purchaser to complete the model usage rights transaction.

5. The multi-party trusted AI model transaction settlement method based on blockchain technology according to claim 4, characterized in that, Step S4: Using asymmetric encryption and ERC721 smart contracts, verify the model purchaser's ownership of the wallet address and the corresponding NFT held by that wallet address. If the verification passes, create a temporary account and issue it to the model purchaser. Specifically, this includes: Step S41: The model purchaser initiates a request for model usage credentials, the request including its wallet address in the blockchain and the NFT's Token ID; Step S42: The model purchaser signs the request using their private key and publishes their public key to the blockchain using their wallet account for signature verification; Step S43: Verify the signature of the request. After successful signature verification, verify with the smart contract whether the wallet address owns the corresponding NFT. Step S44: If the verification is successful, create a temporary account with only pull permissions for the project containing the model, and return it to the model purchaser as a credential for using the model.

6. A multi-party trusted AI model transaction settlement system based on blockchain technology, characterized in that, Includes the following modules: The model image management module is used to process and store model image files, verify users' rights to use the model, and create and manage image usage credentials. The smart contract transaction module is used to process transaction requests, perform balance inquiries, authorizations, and transfers of ERC20 tokens, mint, transfer, and verify NFTs, and store transfer records on the blockchain; The front-end user interface module provides a user-friendly interface for logging into Ethereum wallets and interacting directly with smart contracts.

7. The multi-party trusted AI model transaction settlement system based on blockchain technology according to claim 6, characterized in that, The model image management module includes the following modules: The backend business view unit is used to process business requests sent from the frontend, call the functions of other units in this module, and perform file management and database operations. The container operation unit is used to load uploaded container image files, add tags, and push the image to the image repository; The image repository unit is used to store container images, isolate permissions for different images, and create temporary accounts to pull images.

8. The multi-party trusted AI model transaction settlement system based on blockchain technology according to claim 6, characterized in that, The smart contract transaction module includes the following modules: The ERC20 circulating token management contract is used to issue, query, and transfer ERC20 standard tokens circulating within the system, and to authorize the main business logic contract to operate on the tokens. ERC721 Nonfungible Token Management Contract is used to mint, query, transfer and verify nonfungible tokens representing the right to use the model within the system, and authorize operable NFT tokens for the main business logic contract. IPFS (InterPlanetary File System) units are used to store NFT metadata files and provide CID as the token URI of the NFT pointing to the target; The transaction business logic smart contract is used to handle the main transaction requests, transfer authorized ERC20 tokens and NFTs, and complete the transaction; The blockchain browser unit is used to query token transfer records within the smart contract transaction module, making transaction information publicly and transparently displayed to any user.

9. The multi-party trusted AI model transaction settlement system based on blockchain technology according to claim 6, characterized in that, The front-end user interface module includes the following modules: The view control unit is used for basic page element display and interaction with the model mirror management module; The Ethereum wallet management unit is used to log in to the Ethereum wallet, interact with smart contracts using the Application Binary Interface (ABI), and provides a blockchain-related operation interface.

10. A multi-party trusted AI model transaction settlement device based on blockchain technology, characterized in that, It includes one or more electronic devices, wherein the one or more electronic devices are used to implement the method of any one of claims 1 to 5.

Images