AI model data hosting and transaction integration method

Abstract

The invention discloses an AI model data trusteeship and transaction integration method, which is realized through four stages: an integrated architecture establishment stage: constructing a'trusteeship-transaction fusion layer ', and completing data ownership anchoring and transaction rule embedding; in the trusteeship transaction cooperation stage, data listing linkage, intelligent matching transaction and fund-authority linkage are realized; in the post-transaction authority management and control stage, data use is limited through a data sandbox, authority is dynamically adjusted, and earnings are automatically distributed; in the whole-process tracing stage, trusteeship, transaction and use information is linked to form a data resume, and risks are monitored in real time and positioned and disposed. The method is characterized in that a trusteeship-transaction fusion framework is constructed (state real-time linkage is realized), a use right exclusive transaction and sandbox management and control mode is designed (data abuse is avoided), full-process intelligent contract collaboration is adopted (transaction efficiency is improved), and the method is suitable for the fields of medical treatment, automatic driving, industry and the like which have high requirements on data compliance and transaction safety.

Description

Technical Field

[0001] This invention relates to the intersection of AI data management and digital transactions, specifically to an integrated method for AI model data hosting and trading. Background Technology

[0002] In the current AI field, data hosting and data trading are mostly independent systems, with core issues concentrated in three aspects.

[0003] Firstly, there is a disconnect between custody and trading: In the existing solution, data must first be stored on the custody platform and then listed for trading through a third-party trading platform. The two are not directly linked, which leads to repeated verification of data compliance and ownership information during trading, making the process cumbersome and inefficient. At the same time, after the transaction is completed, the permission settings of the custody platform need to be manually synchronized, which can easily lead to the mismatch problem of "the transaction is completed but the data cannot be accessed".

[0004] Secondly, the transaction model is simplistic and lacks security: most data transactions are "ownership transfer" models, which can easily lead to secondary dissemination or misuse of data; some usage rights transactions rely on the trust between the two parties, lack a hosting platform to control the scope of data use in real time, and cannot protect the rights and interests of data providers.

[0005] Third, there is a lack of coordinated management and control throughout the entire process: the custody stage is only responsible for storage, and the transaction stage is only responsible for matching. There is no integrated coordinated mechanism of "custody - listing - matching - transaction - use - revenue". Changes in ownership, revenue distribution, and risk warnings in data transactions cannot be synchronized with the custody status in real time, which increases the risk of transaction disputes.

[0006] While existing technologies have achieved single functions such as data custody or trading, or explored data assetization, securitization (focusing on value circulation), trust custody (focusing on risk isolation), and version mapping (focusing on collaborative management), they have not designed an integrated architecture for the "strong correlation between custody and trading," nor have they built a collaborative mechanism adapted to AI data where "custody supports trading and trading feeds back into custody." They cannot meet the requirements of "efficient trading, secure and controllable, and closed-loop processes" for data, which is fundamentally different from the four existing patents. Summary of the Invention

[0007] The purpose of this invention is to provide an integrated method for AI model data hosting and trading, aiming to solve the problems of disconnect between AI model data hosting and trading, insufficient security of trading models, and lack of full-process collaborative management in the existing technology. This invention provides an integrated method for AI model data hosting and trading, which realizes real-time linkage between data hosting and trading, secure trading of usage rights, and intelligent management and control of the entire process by constructing a hosting-trading fusion architecture, thereby improving the efficiency and security of data trading.

[0008] To address the aforementioned technical problems, this invention provides a method for integrating AI model data hosting and trading, with the following specific steps: 1. Integrated Architecture Setup Phase Integration Module Deployment: A "Custody-Trading Integration Layer" is constructed as the core, integrating the data custody module and the trading module. The custody module is responsible for encrypted data storage, ownership registration, and compliance verification; the trading module is responsible for listing, intelligent matching, and price matching. The integration layer achieves real-time data synchronization between the two modules through API interfaces, ensuring consistency between the custody data status and the trading status.

[0009] Data ownership anchoring: When the data provider uploads AI model data to the hosting module, the ownership registration contract automatically generates a unique data identifier (format: "Data-Domain-Type-Timestamp"), associates it with the data provider information, ownership certificate documents and usage restrictions (such as prohibition of secondary transactions and usage duration), generates a "Data Ownership Certificate" and writes it to the blockchain as the ownership basis for transactions.

[0010] Embedded transaction rules: Transaction rules are preset in the integrated smart contract, including transaction mode (limited to usage rights transactions), pricing method (fixed price / auction), revenue sharing ratio (revenue sharing between data provider and platform), and compliance review standards (such as data anonymization requirements); the rules take effect after consensus among blockchain nodes and serve as the basis for the execution of subsequent transactions.

[0011] 2. Custody Transaction Collaboration Phase Data listing linkage: When a data provider initiates a listing application in the trading module, the smart contract automatically calls the data source information (data type, data volume, compliance status) of the custody module, eliminating the need for manual uploading; at the same time, it verifies the validity of the "Data Ownership Certificate". If there is a dispute over ownership of the custody data or the compliance verification is not completed, the listing application will be rejected directly, realizing "custody status determines trading qualification".

[0012] Intelligent matching transactions: After receiving the application (including data type, purpose, and budget) from the data requester, the transaction module combines the data tags in the escrow module (such as "medical imaging - lung nodules - annotation accuracy 95%) with the requester's profile, filters suitable data, and pushes it to the requester; after the requester confirms the transaction, the integrated smart contract automatically freezes the transaction permissions of the corresponding data in the escrow module (to avoid duplicate transactions), and generates a "Data Transaction Agreement", which is written to the blockchain after being confirmed by both parties.

[0013] Transaction Fund Management: The requester transfers transaction funds to the co-managed account of the integration layer. After the fund management contract detects the arrival of funds, it sends a "transaction completed" instruction to the custody module. The custody module synchronously updates the data status to "transaction in progress" and generates a temporary authorization key (which can only be used by the requester), realizing the linkage of "access is granted as soon as funds arrive".

[0014] 3. Post-transaction access control phase Data Sandbox Usage: The requester uses data in the data sandbox built into the fusion layer through a temporary authorization key (the original data cannot be downloaded or copied). The sandbox records data usage behavior (such as the number of calls and the type of training model) in real time and synchronizes it to the hosting module. If the usage behavior exceeds the agreement in the "Data Transaction Agreement" (such as for training non-agreement models), the hosting module will immediately freeze the authorization key and terminate the data usage.

[0015] Dynamic permission adjustment: If the requesting party needs to extend the usage period or expand the usage scope, it needs to submit a change application in the transaction module; the current data status of the escrow module is integrated with the smart contract to call the escrow module (whether it has been traded by other parties), and after the review is approved, the "Data Transaction Agreement" is updated, and the authorization permissions and duration of the escrow module are adjusted in sync to achieve "real-time linkage between transaction changes and permission adjustments".

[0016] Automatic revenue distribution: After the transaction is completed (or the usage period expires), the revenue distribution contract will transfer the funds in the co-managed account to the data provider and platform accounts according to the preset sharing ratio; after the distribution is completed, the custody module will automatically update the data status to "tradable", allowing it to be listed again, forming a closed loop of "transaction - distribution - re-transaction".

[0017] 4. Full-process traceability stage Information traceability on the blockchain: By integrating smart contracts, information on the entire data custody process (storage address, ownership change, compliance records) and information on the entire transaction process (listing records, matching results, fund flow, usage logs) are written into the blockchain in real time, forming an immutable "data history" that can be queried at any time by data providers, demanders, and regulators.

[0018] Risk warning and traceability: The risk monitoring contract monitors the status of escrow data (such as whether it has been illegally accessed) and transaction status (such as abnormal fund transfers) in real time. If a risk is detected, an alert is immediately triggered and pushed to the relevant parties. At the same time, the "data history" in the blockchain is called to locate the link where the risk occurred (such as abnormal escrow storage or transaction matching error) to assist in rapid handling.

[0019] In summary, due to the adoption of the above-mentioned technologies, the beneficial effects of this invention are: 1. Unlike existing independent systems, this invention constructs a "custody-transaction integration layer" to achieve real-time linkage between data custody status and transaction eligibility, permission adjustment, and fund management, thus solving the problem of process disconnect.

[0020] 2. Breaking through the traditional ownership transaction model, it only opens up the right to use data, restricts data download and copying through a built-in data sandbox, and combines real-time permission control of the hosting module to prevent data abuse and protect the rights and interests of data providers.

[0021] 3. The innovative design integrates smart contracts, covering the entire process of "custody - listing - matching - trading - usage - revenue", realizing automatic execution of transaction rules, real-time status synchronization, and full on-chain information, reducing manual intervention and lowering the risk of transaction disputes. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention, making other features, objects, and advantages of the invention more apparent. The illustrative embodiments of the invention illustrated in the drawings and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a system framework diagram of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to represent selected embodiments of the invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] In the description of this invention, it should be understood that the terms indicating orientation or positional relationship are based on the orientation or positional relationship shown in the drawings and are only for the convenience of describing the invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention.

[0025] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific context of the specification.

[0026] This invention provides a method for integrating AI model data hosting and trading, comprising: Implementation steps (taking autonomous driving AI training data as an example) Integrated architecture construction An autonomous driving company (data provider) deployed a "hosting-trading fusion layer" that integrates a data hosting module (storing LiDAR point cloud data) and a trading module (supporting data listing); the fusion layer achieves real-time data synchronization between the two modules through API.

[0027] The provider uploads 100,000 LiDAR annotation data points for urban roads, and the ownership registration contract generates the data identifier "Data-Auto-Lidar-202X1210", associates the provider information and usage restrictions (limited to L2 level autonomous driving model training, secondary transactions are prohibited), and generates a "Data Ownership Certificate" which is written to the blockchain.

[0028] The system integrates smart contract preset transaction rules: a usage right transaction model with a fixed unit price of 1 yuan per transaction, revenue sharing (90% for the provider and 10% for the platform), and compliance standards of "de-identification processing + laser point cloud de-identification". The rules take effect through consensus among consortium blockchain nodes.

[0029] Custody Transaction Collaboration The provider initiates a listing application in the trading module, and the smart contract automatically calls the data from the escrow module: the data volume is 100,000 records, the compliance status is "de-identified", and the ownership is undisputed. The application is directly approved and the listing information (data identifier, unit price, and usage restrictions) is generated.

[0030] A new energy vehicle company (the demander) submitted a request: "L2-level autonomous driving LiDAR data, budget of 80,000 yuan, for rainy day scenario model training"; the intelligent matching contract screened out the above-mentioned listing data (matching the demand) and pushed it to the demander.

[0031] The demander confirms the transaction (purchasing 80,000 data entries) and transfers 80,000 yuan to the co-managed account. The fund management contract detects the arrival of funds and sends an instruction to the custody module. The custody module updates the data status to "transaction in progress," generates a temporary authorization key with a 7-day validity period, and pushes it to the demander's terminal.

[0032] Post-transaction access control The requester logs into the fusion layer data sandbox using a temporary key and calls 80,000 data points to train a rainy day scenario model; the sandbox records usage behavior in real time (calling 60,000 data points, training the model to L2 level autonomous driving) and synchronizes it to the managed module.

[0033] The client requested a 3-day extension of the usage period due to training needs and submitted a change request in the transaction module. The integrated smart contract called the escrow module to confirm that the data had not been traded by other parties. After the review was approved, the "Data Transaction Agreement" was updated, and the escrow module extended the validity period of the authorized key to 10 days.

[0034] Upon completion of the 10-day usage period, the revenue distribution agreement allocates funds at a 9:1 ratio: the provider receives 72,000 yuan and the platform receives 8,000 yuan. The funds are transferred from the co-managed account to the corresponding account. The escrow module automatically updates the data status to "tradable," allowing the provider to relist the remaining 20,000 data entries.

[0035] Full process traceability The integrated smart contract writes all data hosting records (storage address, anonymization time), transaction records (listing time, transaction amount), and usage records (number of calls, training model) into the blockchain, forming a data history of "Data-Auto-Lidar-202X1210".

[0036] The risk monitoring contract detects that the demander attempts to copy data in the sandbox, immediately freezes the authorization key, and pushes an alert to the provider and the platform; by locating the "violation in the usage process" through blockchain data history, the platform verifies and terminates the transaction, deducts 10% of the demander's deposit, and protects the provider's rights.

[0037] Example: Using AI diagnostic model data from mammography in the medical field as an application scenario I. Setting up the implementation environment Blockchain node configuration: Six consortium blockchain nodes are set up, held by the hospital, medical AI company, platform operator, health commission regulatory department, medical ethics committee and distributed storage service provider respectively. The consensus mechanism adopts RBFT (response latency ≤500ms) to ensure the compliance and traceability of medical data transactions.

[0038] Core module adaptation: The "Hosting-Transaction Integration Layer" adds a "Medical Data Compliance Sub-module" (connecting to the verification rules of the "Medical Data Security Guidelines") and an "Ethics Review Interface" (linking with the Medical Ethics Committee node); the data hosting module integrates a "Medical Image Desensitization Engine" (automatically masking patient names, medical record numbers, and other identifying information); the data sandbox has a built-in "Medical Model Training Interface" (only supporting mammography X-ray diagnostic model training, other uses are prohibited).

[0039] Terminal and storage setup: The hospital is equipped with medical data acquisition terminals (connected to the PACS system, automatically filtering and labeling compliant data), the AI ​​company is equipped with medical AI R&D terminals (registered with the National Health Commission), and the storage system adopts medical-grade distributed cloud storage (compliant with the "Health and Medical Big Data Storage Standard"), which is connected to the fusion layer through a medical-grade encrypted SDK, supporting real-time data upload and authorized access. Specific implementation steps

[0040] (I) Integrated Architecture Setup Phase Deployment of the integration module: The platform operator builds a "hosting-transaction integration layer" that integrates a medical data hosting module (storing mammogram data) and a transaction module (only open to registered medical AI institutions). The two modules achieve real-time data synchronization through medical-grade APIs to ensure that the compliance status of the hosted data is linked to the transaction eligibility.

[0041] Data ownership anchoring: The hospital uploads 50,000 breast X-ray labeled data (already anonymized), and the ownership registration contract generates a unique data identifier "Data-Med-Mammo-202X0115" (format: Data-domain-data type-timestamp), which is associated with the hospital name, the "Medical Data Use Authorization Letter" and usage restrictions (limited to breast AI diagnostic model training, usage period of 30 days), and generates a "Medical Data Ownership Certificate", which is written to the blockchain after being reviewed by the health commission's regulatory node.

[0042] Embedded transaction rules: The smart contract pre-sets the medical data transaction rules: The transaction mode is "exclusive right of use transaction" (only one demand party is authorized at the same time), the pricing method is a fixed price of 0.8 yuan / data, the revenue sharing ratio is (85% for hospitals and 15% for platforms), and the compliance standard is "compliance with Article 5.2 of the Medical Data Security Guidelines (de-identification requirements) + passing ethical review"; the rules take effect after consensus is reached by all alliance chain nodes.

[0043] (II) Custody Transaction Coordination Phase Data listing linkage: When a hospital initiates a listing application in the transaction module, the smart contract automatically calls the data information from the hosting module: 50,000 data entries, the anonymization status is "completed (covering the identification information)", the ownership certificate is "valid", and the ethics review is "passed". The listing is directly approved and the listing information (data identification, unit price, usage restrictions, and compliance certificate) is generated.

[0044] Intelligent matching transaction: A medical AI company submits a request: "40,000 breast X-ray labeled data entries, to optimize the malignant lesion recognition rate of the AI ​​model, with a budget of 32,000 yuan"; The intelligent matching contract combines the data tags of the escrow module ("breast X-ray - benign / malignant labeling - accuracy 92%) with the profile of the requester ("registered medical AI institution - specializing in breast diagnostic models") to filter out the above-mentioned listed data and push it to the AI ​​company.

[0045] Transaction fund management: After the AI ​​company confirms the transaction, it will transfer 32,000 yuan to the medical-specific co-management account of the integration layer (supervised by the National Health Commission); the fund management contract will detect the arrival of funds and send a "transaction completed" instruction to the escrow module. The escrow module will immediately update the status of 40,000 data entries to "transaction in progress (exclusive use)" and the remaining 10,000 entries to "tradeable". At the same time, it will generate a temporary authorization key with a validity period of 30 days and push it to the AI ​​company's R&D terminal.

[0046] (III) Post-Transaction Access Control Phase Data Sandbox Usage: The AI ​​company logs into the fusion layer medical data sandbox using a temporary key and calls 40,000 data points to train the model; the sandbox records usage behavior in real time (20,000 data points were called on January 20, with the training scenario being "early malignant lesion identification"), and synchronizes it to the hosting module and the health commission's regulatory node; when the AI ​​company attempts to download raw mammogram images, the sandbox immediately triggers an interception, sends a "violation warning" to the hosting module, and the hosting module temporarily freezes the authorization key, restoring access after the AI ​​company confirms "use only within the sandbox".

[0047] Dynamic permission adjustment: Due to the extended model training cycle, the AI ​​company applied to extend the usage period by 10 days and submitted a change application in the transaction module; the integrated smart contract called the hosting module to confirm that the 40,000 data entries had not been reserved by other parties and that the AI ​​company had no violation records. After the review was approved, the "Medical Data Transaction Agreement" was updated, and the hosting module extended the validity period of the authorization key to 40 days and simultaneously reported to the health commission's regulatory node.

[0048] Automatic revenue distribution: After the 40-day usage period, the AI ​​company confirms that the data usage has been completed, and the revenue distribution contract allocates funds in an 85:15 ratio: the hospital receives 32,000 yuan × 85% = 27,200 yuan, and the platform receives 32,000 yuan × 15% = 4,800 yuan. The funds are transferred from the co-managed account to the hospital's dedicated medical data revenue account (misappropriation is prohibited); the hosting module automatically updates the status of 40,000 data entries to "tradable", allowing the hospital to relist the data.

[0049] (iv) Full-process traceability stage Information traceability on the blockchain: The integration of smart contracts writes all process information into the blockchain: escrow records (data storage address, anonymization time, ethical review opinions), transaction records (listing time, transaction amount, fund arrival time), and usage records (number of calls, training scenarios, violation interception records), forming a medical data history of "Data-Med-Mammo-202X0115", which can be queried at any time by hospitals, AI companies, and health commissions.

[0050] Risk warning and traceability: The risk monitoring contract detected an unregistered IP address attempting to access the sandbox, immediately triggering a level-two warning and pushing an "illegal access notification" (including IP address and access time) to all consortium blockchain nodes; through blockchain data history, the "sandbox access process abnormality" was located, and the platform, in conjunction with regulatory authorities, verified that it was due to an AI company employee's misoperation (using an unregistered terminal), and urged them to replace it with a registered terminal to restore access, without causing data leakage.

Claims

1. A method for integrating AI model data hosting and trading, characterized in that: Includes the following steps: (1) Integrated architecture construction phase: Construct a "custody-trading integration layer" to integrate the data custody module and the trading module, and achieve real-time data synchronization between the two modules; When a data provider uploads data, the ownership registration contract generates a unique data identifier and a "Data Ownership Certificate," which are then written into the blockchain. Transaction rules are preset in the integrated smart contract and take effect after consensus among blockchain nodes; (2) Custody and Transaction Coordination Phase: When data is listed, the integrated smart contract calls the escrow module to verify the data's compliance and ownership. Once verified, the listing information is generated. The transaction module receives applications from demanders, intelligently matches contracts, combines escrow data tags with demand profiles to filter suitable data, and freezes data transaction permissions after a transaction is completed. When the demander's funds are transferred into the co-managed account, the fund management contract triggers the custody module to generate a temporary authorization key, thus granting data access permissions. (3) Post-transaction access control phase: The requester uses data in the data sandbox of the fusion layer. The sandbox records the usage behavior and synchronizes it to the managed module. If there is a violation, the user's access will be frozen. When a transaction changes, the smart contract reviews the status of the escrow data, and updates the protocol and adjusts the authorization permissions after approval. After the transaction is completed, the profit distribution contract will transfer funds according to the preset ratio, and the custody module will update the data status to be tradable; (4) Full-process traceability stage: By integrating smart contracts, information about the entire process of custody, transaction, and use is written into the blockchain, forming a data history. Risk monitoring contracts detect anomalies, triggering alerts and locating risk points through data history to assist in handling them.

2. The method for integrating AI model data hosting and trading according to claim 1, characterized in that, The data identifier format is "Data-Domain-Type-Timestamp", where the domain includes autonomous driving, medical, and industrial, and the type includes raw data, labeled data, and derived data.

3. The method for integrating AI model data hosting and trading according to claim 1, characterized in that, The transaction rules include the transaction mode (limited to usage rights transactions), pricing method (fixed price / auction), revenue sharing ratio, and data compliance standards.

4. The method for integrating AI model data hosting and trading according to claim 1, characterized in that, The data sandbox supports data access and model training, but restricts data download, copying, and external transmission, and records usage behavior logs in real time.

5. The method for integrating AI model data hosting and trading according to claim 1, characterized in that, The blockchain platform is either the Hyperledger Fabric consortium blockchain or the FISCOBCOS consortium blockchain, and the smart contracts are written in Go or Solidity.

6. The method for integrating AI model data hosting and trading according to claim 1, characterized in that, The encryption method of the managed module adopts AES-256 or SM4 algorithm, and the data storage adopts IPFS distributed storage system.

7. The method for integrating AI model data hosting and trading according to claim 1, characterized in that, The anomalies monitored by the risk monitoring contract include illegal data copying, use beyond the agreed scope, abnormal fund transfers, and ownership disputes.

Images