A large model data supervision method and system based on a blockchain

Abstract

The application discloses a kind of big model data supervision method and system based on block chain, it is related to data supervision technical field, including the following steps: receiving big model data, data desensitization and encryption operation are carried out to big model data, obtain the big model data after encryption;The encrypted big model data is input into the control variable grouping small model established in advance to carry out the effect test of privacy protection, obtain the small model training data of qualified test, wherein, by the sampling test process of pre-established control variable grouping small model of the preset consensus node is involved;Based on the small model training data of qualified test obtained is input into big model and is trained, obtain the data after training, based on smart contract and incentive mechanism algorithm, the data after training is punished and identified, and data supervision result is obtained.

Description

Technical Field

[0001] This invention relates to the field of data supervision technology, specifically a large-scale data supervision method and system based on blockchain. Background Technology

[0002] With the rapid development of information technology, big data has become a key force driving technological revolution and social progress. Large models (LMs), with their superior data processing and deep learning capabilities, are gradually changing the way we understand and utilize data, becoming a focal point in the field of artificial intelligence. Large models typically refer to AI models with a large number of training parameters and powerful computing capabilities, capable of handling complex large-scale dataset analysis tasks and providing accurate predictions and decision support. These models have shown enormous potential in fields such as natural language processing, image recognition, and speech recognition, providing a powerful impetus for the digital transformation of various industries. However, the rapid development and widespread application of large model technology have also brought a series of regulatory challenges. The most prominent is data security. The training and application of large models involve the processing of massive amounts of data, which often contains sensitive information. How to ensure the privacy and security of this data during collection, processing, and use has become an urgent problem to be solved. Toxic data is also a potential factor endangering the security of large models. Toxic data refers to data where specific inputs are given in the training data, causing the large model to be built based on incorrect information. The complexity and opacity of large models make it difficult for regulatory agencies to effectively oversee their decision-making processes, and these oversights inevitably pose potential risks of bias and discrimination. Content trained on poorly oriented data may reinforce human biases, thereby solidifying the data used to generate subsequent content. Furthermore, the lack of transparency and interpretability in the machine learning process during the generation of large models poses a risk of collapse to the overall system's robustness and the security of network interactions. Summary of the Invention

[0003] To address the shortcomings mentioned in the background section, the present invention aims to provide a method and system for monitoring large-scale model data based on blockchain.

[0004] Firstly, the objective of this invention can be achieved through the following technical solution: a large-scale model data supervision method based on blockchain, the method comprising the following steps:

[0005] Receive large model data, perform data desensitization and encryption operations on the large model data, and obtain encrypted large model data;

[0006] The encrypted large model data is input into a pre-established small model with controlled variables to test the effectiveness of privacy protection, and the qualified small model training data is obtained. The sampling test process of the pre-established small model with controlled variables is carried out through a preset consensus node.

[0007] The qualified small model training data is input into the large model for training to obtain trained data. Based on smart contracts and incentive mechanism algorithms, the trained data is used to determine rewards and penalties, and the data supervision results are obtained.

[0008] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: the data desensitization process includes: data replacement, data generalization and data perturbation, after which desensitized data is obtained, and the desensitized data is encrypted and protected by using SM9 national cryptographic implicit signature, symmetric encryption algorithm or asymmetric encryption algorithm, and finally the encrypted large model data is obtained.

[0009] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: the desensitized and encrypted data, based on the immutability and transparency of the blockchain, records the access, modification, and processing of data throughout the process, generating verifiable audit and regulatory logs; combining smart contracts, using role-based access control (RBAC) or attribute-based access control (ABAC) to manage data access permissions and implement a data access control mechanism, wherein only authorized users or nodes can access the encrypted data; using digital identity identifiers on the blockchain for identity authentication and authorization, designing a blockchain-based identity authentication mechanism; large model training data, after privacy protection, participates in the training of the large model; combining an incentive mechanism, the trained data is sampled and analyzed; if privacy is still exposed, the nodes and organizations involved in the audit can be traced according to the logs, and rewards and punishments can be applied.

[0010] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: the preset consensus node uses a consensus mechanism to judge the authenticity and keyword bias of data in the large model, and the effectiveness of the ability to judge the authenticity and bias of data is tested by designing experiments to verify the accuracy of the algorithm in the blockchain system. By using real datasets or synthetic datasets to conduct controlled variable grouping experiments, the authenticity and bias of data under different scenarios are simulated, the obtained analysis experimental results are optimized, the performance of the system in terms of data authenticity and bias is analyzed based on the experimental results, and the algorithm and system design are optimized in a targeted manner to form positive feedback, thereby performing data compliance judgment.

[0011] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: when determining rewards and punishments for trained data based on smart contracts and incentive mechanism algorithms, the data subject is incentivized to participate in data supervision contributions to the maximum extent, allowing the data supervisor to receive corresponding compensation, while competition eliminates low-quality data providers, thus forming a supervision and punishment mechanism.

[0012] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: the supervision and punishment mechanism process is as follows:

[0013] Rewards and penalties are recorded on the blockchain. Every reward or penalty transaction is recorded on the blockchain and can be viewed and verified by all participants. The incentive mechanism is managed in conjunction with smart contracts. The execution logic of the incentive mechanism is managed by blockchain smart contracts. The smart contracts automatically execute reward and penalty operations according to preset rules and conditions, and the fully open and transparent mechanism ensures that all participants understand the execution standards of rewards and penalties.

[0014] The execution logic in smart contracts takes into account data privacy issues. On-chain nodes or clients vote and govern on-chain, and combined with task evaluation algorithms, rewards are dynamically adjusted based on indicators such as task difficulty, completion, and quality. The concept of proof-of-work is adopted, and the reward distribution ratio is determined based on the actual contribution of the task. Market participants jointly participate in the formulation and adjustment of incentive mechanism rules. After participating in the consensus phase, an on-chain voting mechanism is designed to determine the reward distribution method or the degree of punishment.

[0015] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: the smart contract participating in the design of an incentive mechanism, and in cooperation with the consensus mechanism, designing a reward mechanism based on data compliance and contribution to encourage users to provide compliant data and participate in data processing, rewarding compliant data providers and penalizing non-compliant or consensus-abnormal processors; through the data structure and logic construction of the smart contract, including data storage, processing, and access control, defining data attributes in the contract, leveraging the characteristic of no human intervention, giving full play to the advantages of smart contracts, and using smart contracts to construct a small model from privacy-processed data, which is a large model with limited training volume and scale, built based on pre-trained data and some compliant data. The data participating in the training of the small model is divided into three groups: all pre-trained data; trained qualified data; and pre-trained data. Half of the data is qualified training data. Then, the small model undergoes routine testing, defining standards and rules for compliant data. If the small model's performance metrics are qualified under the group experiment, the pre-training results are directly added to the large model as compliant data. Conversely, data that does not meet the compliance standards is identified and processed, and non-compliant data is eliminated. An incentive mechanism is used to reward or punish the responsible parties. At the level of designing data bias detection algorithms, algorithms are designed to assess data bias based on data attributes and characteristics, and bias detection logic. The distributed nature of blockchain is used to evaluate data bias and record the results. The logic and algorithm implementation of smart contracts are continuously optimized based on the test results. Relying on the characteristics of smart contracts, small models based on smart contracts are independently established, or incentive mechanisms are combined to promote the contribution and processing of compliant data. Finally, a large model data supervision system is obtained by combining blockchain.

[0016] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: the large-scale model data supervision system is divided into an access layer, a data layer, a network layer, a consensus layer, a contract layer, and an application layer. The chain members in this supervision blockchain are divided into large-scale model data institutions and computer organizations. Based on their business categories, they respectively act as data aggregators and internal algorithm providers, executing corresponding functional requirements. The access layer is the program call layer, which, through interface with the consortium blockchain business interface, executes corresponding algorithms for the large-scale model data and performs supervision. The data layer embodies the classic chain-like data structure of blockchain, mainly consisting of the previous block hash (Pre-Hash) and timestamps. The system consists of a random number (Nonce), MerkleRoot, block height, and a large-scale model data supervision block. Within the network layer, distributed nodes use encryption algorithms to authenticate and control the identities of participating nodes during encrypted data transmission. At the consensus layer, a benchmark consensus algorithm is used to establish the consensus foundation within the system, ensuring information consistency for each node in the blockchain. The contract layer integrates small models, and automatically executed code can quickly respond to compliance checks and access permission requirements. Various smart contracts are used to implement conditional constraints during the large-scale model data supervision process using specific scripts. The application layer is used to construct incentive mechanisms.

[0017] Secondly, in order to achieve the above objectives, this invention discloses a large-scale model data supervision system based on blockchain, comprising:

[0018] The data encryption module is used to receive large model data, perform data desensitization and encryption operations on the large model data, and obtain encrypted large model data.

[0019] The data verification module is used to input the encrypted large model data into the pre-established small control variable group model to verify the effectiveness of privacy protection, and obtain qualified small model training data. The module participates in the sampling verification process of the pre-established small control variable group model through a preset consensus node.

[0020] The data incentive module is used to input the qualified small model training data into the large model for training, obtain the trained data, and determine the reward or punishment of the trained data based on smart contracts and incentive mechanism algorithms to obtain the data supervision results.

[0021] In another aspect of the present invention, in order to achieve the above objectives, a terminal device is disclosed, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. The memory stores the computer program capable of running on the processor, and when the processor loads and executes the computer program, it employs a blockchain-based large-scale data supervision method as described above.

[0022] The beneficial effects of this invention are:

[0023] This invention analyzes existing data supervision mechanisms and proposes a design scheme for a large-scale model data supervision system based on blockchain technology. It relies on four independent and collaborative modules: privacy protection, consensus algorithm, incentive mechanism, and smart contracts to conduct end-to-end data supervision of large-scale model data, from raw metadata to input into the large-scale model, thus rationally utilizing blockchain technology to form a large-scale model data supervision system. Simultaneously, a model of the large-scale model data supervision system based on blockchain technology is established. By analyzing the relationship between blockchain and large-scale model data supervision, the mechanism of large-scale model data supervision based on blockchain technologies such as privacy protection, consensus algorithm, incentive mechanism, and smart contracts is explained, providing insights and suggestions for reasonable, legal, and effective supervision. The proposed supervision system design not only improves the trust level of data management but also achieves automated compliance checks and regulatory reports through smart contracts, reducing the risk of human intervention. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart of the method of the present invention;

[0026] Figure 2 This is a framework diagram of the method system of the present invention;

[0027] Figure 3 This is a module framework diagram of the large model data privacy protection scheme in this invention;

[0028] Figure 4 This is a framework diagram of the large model data consensus algorithm module in this invention;

[0029] Figure 5 This is a framework diagram of the large model data incentive mechanism module in this invention;

[0030] Figure 6 This is a framework diagram of the large model data smart contract module in this invention;

[0031] Figure 7 This is a diagram of the large-scale model data supervision system architecture based on blockchain technology in this invention;

[0032] Figure 8 This is a simplified diagram illustrating the execution flow of the large model data supervision system in this invention;

[0033] Figure 9 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0034] 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 some embodiments of the present invention, and not all embodiments. 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.

[0035] Example 1:

[0036] The following is a description of the relevant terms used in the embodiments of this application:

[0037] Blockchain (also known as blockchain or blockchain chain) is a decentralized distributed ledger that is stored in blocks, is immutable, secure, and reliable. It combines distributed storage, peer-to-peer transmission, consensus mechanisms, and cryptography to record transactions and information through a continuously growing chain of data blocks, ensuring data security and transparency. Blockchain originated with Bitcoin. From its inception, the Bitcoin network has evolved into a global technology, attracting global attention and investment. Subsequently, the emergence of next-generation blockchain platforms such as Ethereum further expanded its application areas. The characteristics of blockchain include decentralization, immutability, transparency, security, and programmability. Each data block is linked to the previous block, forming a continuous chain that ensures the integrity of transaction history. Smart contract technology makes blockchain programmable, supporting a wider range of applications. Blockchain is widely used in finance, supply chain, healthcare, real estate, and other fields. Although it still faces scalability and regulatory challenges, it has become a powerful tool for changing traditional business and social models and has enormous potential for the future.

[0038] like Figure 1 As shown, a blockchain-based method for regulating large-scale model data includes the following steps:

[0039] S101: Receive large model data, perform data desensitization and encryption operations on the large model data, and obtain encrypted large model data;

[0040] From the very beginning of data collection, the data used for training large models requires a series of privacy protection measures. This invention's blockchain-based privacy protection method first considers data anonymization, de-identifying sensitive personal and corporate information within the data and hiding its specific content or key features. Data anonymization is achieved through methods such as data replacement, data generalization, and data perturbation to protect data privacy. The anonymized data then undergoes encryption protection using methods such as SM9 implicit signatures, symmetric encryption algorithms (such as AES), or asymmetric encryption algorithms (such as RSA and ECC) to ensure that the data used for training large models cannot have its metadata reversed. The large model data is distributed and stored across multiple nodes in the blockchain network to ensure data decentralization and security. Simultaneously, data security during storage and transmission must be ensured. The immutability and distributed nature of blockchain enhance data security and trustworthiness. Benefiting from the immutability and transparency of blockchain, the anonymized and encrypted data is recorded throughout the process, generating verifiable audit and regulatory logs to ensure data compliance and security. By combining smart contracts, role-based access control (RBAC) or attribute-based access control (ABAC) can be used to manage data access permissions, implementing a data access control mechanism that allows only authorized users or nodes to access encrypted data. Digital identities on the blockchain are used for authentication and authorization, designing a blockchain-based identity authentication mechanism to ensure the legitimacy and trustworthiness of users or nodes participating in data access and processing. Large-scale model training data, after privacy protection, is used for training the large model. An incentive mechanism is used to sample and analyze the trained data; if privacy is still exposed, the nodes and organizations involved in the audit can be traced based on logs, and rewards and penalties can be imposed accordingly. The sub-modules of the measures described in this example and the overall privacy protection module are attached. Figure 3 As shown.

[0041] S102: Input the encrypted large model data into the pre-established small model with controlled variables to test the effectiveness of privacy protection, and obtain qualified small model training data. The sampling test process of the pre-established small model with controlled variables is carried out through the preset consensus node.

[0042] When using consensus mechanisms in pre-defined consensus nodes to determine the authenticity and keyword bias of data in large models to improve data regulatory compliance, the following measures can be taken. First, choosing a suitable consensus mechanism in a blockchain system is crucial for the stability and reliability of the entire system. Using Proof of Stake (PoS) or Delegated Proof of Stake (DPoS), which are energy-efficient, highly efficient, and fast in transaction speeds, as benchmarks is the cornerstone for ensuring the smooth operation of the system. With a robust consensus mechanism as support, the source of the data becomes the first line of defense when verifying its authenticity. Before the data is added to the large model training, technologies such as digital signatures and cryptographic hashes can be used to ensure the integrity and authenticity of the data. Recording the unique identifier of the data in the blockchain, along with the history of data changes, ensures the traceability and immutability of the blockchain throughout the entire process, including data creation, modification, and deletion. For regulatory measures to detect keyword bias, a data bias detection algorithm needs to be designed, with the algorithm setting specific indicators for data bias detection based on the specific problem. For example, in machine learning models, a fairness-based algorithm can be considered to detect data bias. Distributed computing is used for data detection, distributing the bias detection task across multiple blockchain nodes for parallel computation. The results are recorded on the blockchain, and a consensus mechanism ensures consistency. This algorithm also requires smart contracts to execute the logic of data authenticity verification and bias detection. Smart contracts can serve as part of the consensus mechanism, executing the data verification and detection logic and recording the results on the blockchain. Simultaneously, data access control can be implemented, ensuring that only authorized users or nodes can access and perform data detection operations. The effectiveness of the algorithm in detecting data authenticity and bias requires experimental design to verify its accuracy within the blockchain system. This can be achieved through controlled variable grouping experiments using real or synthetic datasets to simulate data authenticity and bias detection under different scenarios. The experimental results are then analyzed to optimize the system's performance in data authenticity and bias detection, allowing for targeted optimization of the algorithm and system design, creating positive feedback, and continuously improving data compliance judgment. Simultaneously, small-scale data experiments are conducted using internal smart contract models. Compliant training data from these small models is directly incorporated into the larger model, while non-compliant data is discarded. Compliance is also linked to an incentive mechanism. Data security is combined with privacy protection to ensure the security of the blockchain system. Measures such as encrypted storage and secure transmission of data are considered to prevent data leaks and malicious attacks. User and data privacy are considered in the design, adhering to relevant privacy laws and regulations to ensure the legality and compliance of data processing.By combining the above approaches, a data authenticity and bias detection module based on a consensus mechanism can be established. Simultaneously, in practice, issues such as the security, performance, and scalability of the blockchain system should be fully considered, and the system design and algorithm implementation should be continuously optimized and improved. The overall consensus algorithm module described in this example is attached. Figure 4 As shown.

[0043] S103: Based on the qualified small model training data obtained, input it into the large model for training to obtain trained data. Based on smart contracts and incentive mechanism algorithms, reward and punishment are determined on the trained data to obtain data supervision results.

[0044] The method described in this invention establishes a fair and reasonable large-scale model data rights assessment mechanism, maximizing incentives for data subjects to participate in data supervision contributions, ensuring data supervisors receive corresponding rewards, and simultaneously eliminating low-quality data providers through competition, gradually forming a supervision and punishment mechanism. Combining the incentive mechanism with blockchain technology to support large-scale model data supervision can enhance the transparency, credibility, and security of the data market, ensure the fairness and effectiveness of the incentive process, and increase the activity of data supervision. Specifically, rewards and punishments are recorded and stored on the blockchain, ensuring the immutability and transparency of the records. Every reward or punishment transaction is recorded on the blockchain and can be viewed and verified by all participants, thereby enhancing the credibility of the data market. The incentive mechanism is managed through linkage with smart contracts. Blockchain smart contracts are used to manage the execution logic of the incentive mechanism, ensuring the automated execution of rewards and punishments. Smart contracts can automatically execute reward and punishment operations according to preset rules and conditions, reducing human intervention and errors, and the fully transparent mechanism ensures that all participants understand the execution standards for rewards and punishments. At the same time, the execution logic in the smart contracts also considers data privacy issues to avoid the leakage of sensitive information. On-chain nodes or clients perform on-chain voting and governance, dynamically adjusting rewards based on task evaluation algorithms and metrics such as task difficulty, completion, and quality. Employing a concept similar to Proof of Work, the reward distribution ratio is determined based on the actual contribution of the task, avoiding a one-size-fits-all reward mechanism. Market participants are involved in the rule-making and adjustment of the incentive mechanism. After participating in the consensus phase, an on-chain voting mechanism is designed to determine the reward distribution method or the severity of penalties. Each reward distribution is recorded via smart contracts to ensure the authenticity and traceability of the rewards. Simultaneously, data security and privacy protection are paramount; data security and privacy protection must be considered in reward and penalty records. It needs to be integrated with a privacy protection module, using encryption technology to protect sensitive information and ensure that data is not accessed without authorization. Combining the incentive mechanism with blockchain can improve market transparency, fairness, and security, promoting active participation and compliant behavior. The overall module of the incentive mechanism described in this example is attached. Figure 5 As shown.

[0045] The system described in this invention is linked to smart contracts in privacy protection, consensus, and incentive mechanisms. For example, it participates in the design of incentive mechanisms and, in conjunction with consensus mechanisms, designs reward mechanisms based on data compliance and contribution, encouraging users to provide compliant data and participate in data processing, rewarding compliant data providers and penalizing non-compliant or consensus-abnormal processors. These mechanisms rely on well-designed smart contract logic. Through the data structure and logical construction of smart contracts, including functions such as data storage, processing, and access control, data attributes are defined in the contract, such as data characteristics, data authenticity identifiers, and data bias scores. Multiple features are implemented simultaneously, leveraging the characteristic of requiring no human intervention to fully utilize the advantages of smart contracts and significantly achieve intelligent supervision of large-scale model data. A small model is constructed using smart contracts from privacy-processed data. This model is built based on pre-trained data and some compliant data, thus limiting the training volume and scale of the large model. The data used for training the small model is divided into three groups: all pre-trained data; half pre-trained qualified data and half qualified training data; and all qualified training data. Then, routine testing is conducted on the small model to detect data degradation, authenticity, and bias, defining standards and rules for compliant data, such as authenticity scores and bias scores. If the small model's performance metrics are satisfactory under the grouped experiment, the pre-training results are directly added to the large model as compliant data. Conversely, data that does not meet compliance standards is identified and processed, and non-compliant data is discarded. An incentive mechanism is used to reward or punish the responsible parties. The data degradation and authenticity detection steps can be implemented with specific algorithms and data processing logic, such as using methods like adding noise or perturbation for data degradation, while algorithms are designed to verify data authenticity. At the data bias detection algorithm design level, based on data attributes and characteristics and bias detection logic, an algorithm is designed to assess data bias, and the distributed nature of blockchain is used to evaluate and record the results. Finally, the logic and algorithm implementation of the smart contract are continuously optimized based on the test results to improve the system's stability and efficiency. Leveraging the characteristics of smart contracts, a small model based on smart contracts can be independently built or collaborate with other modules to achieve data degradation, authenticity, and bias detection, as well as the addition of compliant data and the elimination of non-compliant data. Incentive mechanisms are then used to promote the contribution and processing of compliant data, and a harmonious large-scale model data supervision system is constructed using blockchain technology. The various modules and the overall smart contract module described in this example are attached. Figure 6 As shown.

[0046] This invention addresses challenges in large-scale data models, including data privacy risks, data non-compliance, low regulatory activity, and lagging intelligent regulation. It proposes a large-scale data supervision system based on a consortium blockchain. The prototype architecture framework is divided into an access layer, data layer, network layer, consensus layer, contract layer, and application layer. Chain members in this supervision blockchain are divided into large-scale data institutions and computer organizations. Based on their business categories, they respectively act as data aggregators and internal algorithm providers, executing corresponding functional requirements. The access layer is the program invocation layer, where it executes corresponding algorithms for large-scale data and performs supervision through interface with the consortium blockchain's business interface. The data layer embodies the classic chain-like data structure of blockchain, mainly composed of the previous block hash (Pre-Hash), timestamp, nonce, MerkleRoot, block height, and large-scale data supervision block. Its distributed nature ensures that the ledger collectively stores regulatory information. Within the network layer, distributed nodes use encryption algorithms to authenticate and control the identities of participating nodes during encrypted data transmission, providing reliable guarantees for data transmission. The consensus layer employs a benchmark consensus algorithm to establish the foundation for consensus within the system, ensuring information consistency among all nodes in the blockchain and enabling efficient execution of consensus requirements such as distributed systems, compliance, and small-scale model simulation compliance. The contract layer integrates the crucial small-scale model structure; automatically executed code at the contract layer can rapidly respond to compliance checks and access permission requests, significantly improving efficiency. The most important component is various smart contracts, which serve to implement conditional constraints during the large-scale model data supervision process using specific scripts. The application layer primarily serves to construct incentive mechanisms, enhancing the overall data supervision capabilities of the system. The prototype architecture framework described in this example is attached. Figure 7 As shown.

[0047] The operational flow and data flow of the large model data to be trained in the scheme described in this invention are achieved through the cooperation of the blockchain network and off-chain linkage. First, the pre-training data flows into the privacy protection module for data anonymization and encryption. At the same time, the privacy protection module performs corresponding data access control and identity authentication permission allocation for nodes according to the operational flow of the smart contract. Then, the encrypted data flows into the smart contract module, where small models are grouped according to the control variables constructed by the smart contract to verify the effectiveness of privacy protection. The consensus algorithm module and the smart contract module synchronously perform bidirectional anchoring of the operational flow and data flow. Consensus nodes participate in small model sampling and testing. If the small model training data passes the test, it will directly flow into the large model; otherwise, it will be discarded. In the incentive mechanism module, rewards and penalties are determined based on the workload of the nodes. Incentives, consensus, and privacy protection are all bidirectionally controlled, while the smart contract participates unidirectionally, directly controlling the automatic execution of the incentive mechanism algorithm, thereby ensuring the fairness of the process without human intervention. The entire system reasonably realizes orderly supervision of large model data in terms of data privacy protection, data quality compliance, improved data supervision activity, and intelligence. In this example, the operational flow and data direction of the large model data to be trained entering this regulatory system are achieved through the cooperation of the blockchain network and off-chain linkage. The operational steps are attached. Figure 8 As shown.

[0048] Example 2: Second aspect, such as Figure 9 As shown, in order to achieve the above objectives, this invention discloses a large-scale model data supervision system based on blockchain, comprising:

[0049] Data encryption module 11 is used to receive large model data, perform data desensitization and encryption operations on the large model data, and obtain encrypted large model data;

[0050] The data verification module 12 is used to input the encrypted large model data into the pre-established control variable grouped small model to verify the effectiveness of privacy protection, and obtain qualified small model training data. The module participates in the sampling verification process of the pre-established control variable grouped small model through a preset consensus node.

[0051] The data incentive module 13 is used to input the qualified small model training data into the large model for training, obtain the trained data, and determine the reward and punishment of the trained data based on smart contracts and incentive mechanism algorithms to obtain the data supervision results.

[0052] Based on the same inventive concept, this invention also provides a computer device, comprising: one or more processors, and a memory for storing one or more computer programs; the programs include program instructions, and the processor executes the program instructions stored in the memory. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, used to implement one or more instructions, specifically for loading and executing one or more instructions stored in a computer storage medium to implement the above-described method.

[0053] It should be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, performs the above-described method. This storage medium can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0054] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0055] The foregoing has shown and described the basic principles, main features, and advantages of this disclosure. Those skilled in the art should understand that this disclosure is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this disclosure. Various changes and modifications can be made to this disclosure without departing from its spirit and scope, and all such changes and modifications fall within the scope of this disclosure as claimed.

Claims

1. A blockchain-based large model data supervision method, characterized in that, The method comprises the following steps: Receiving large model data, data desensitization and encryption operation are performed on the large model data to obtain encrypted large model data; The encrypted large model data is input into a pre-established control variable grouping small model for privacy protection effectiveness test, and qualified small model training data is obtained, wherein a preset consensus node participates in the sampling test process of the pre-established control variable grouping small model; Based on the obtained qualified small model training data, training is performed in the large model to obtain trained data, and based on the smart contract and the incentive mechanism algorithm, rewards and punishments are identified for the trained data to obtain data supervision results; The supervision and punishment mechanism process is as follows: Reward and punishment records are stored on the blockchain, and each reward or punishment transaction is recorded on the blockchain and viewed and verified by all participants; the incentive mechanism is managed through linkage with the smart contract, the execution logic of the incentive mechanism is managed by the blockchain smart contract, the smart contract automatically executes the reward and punishment operations according to the preset rules and conditions, and the fully open and transparent mechanism ensures that all participants understand the execution standards of rewards and punishments; The execution logic in the smart contract considers the problem of data privacy, the on-chain nodes or customers perform on-chain voting and governance, combine the task evaluation algorithm, dynamically adjust the reward according to the task difficulty, completion degree and quality index, adopt the concept of proof of work, determine the allocation proportion of the reward according to the actual contribution of the task, let the market participants jointly participate in the rule making and adjustment of the incentive mechanism, after participating in the consensus stage, design the on-chain voting mechanism to determine the allocation method of the reward or the degree of punishment; The smart contract participates in the design of the incentive mechanism, and in cooperation with the consensus mechanism, an incentive mechanism is designed according to the compliance and contribution of data, to encourage users to provide compliant data and participate in data processing, to reward compliant data providers and punish non-compliant and consensus abnormal processors; through the data structure and logic structure of the smart contract, including data storage, processing and permission control, the attributes of the data are defined in the contract, the advantages of the smart contract are utilized, the data after privacy processing is used to build a small model by the smart contract, and the data participating in the training of the small model is divided into three groups, which are all pre-training data, half of the pre-training data and qualified data, and all qualified training data; then the small model is tested regularly, the standard and rule of compliant data are defined, if the performance index of the small model in the grouping test is qualified, the pre-training result is directly added to the large model as compliant data, otherwise, the data that does not meet the compliance standard is identified and processed, and the non-compliant data is eliminated, and the corresponding responsible party is rewarded and punished in combination with the incentive mechanism, in the design of the data bias detection algorithm, according to the data attributes and characteristics and the bias detection logic, an algorithm is designed to evaluate the bias of the data, and the bias of the data is evaluated and the results are recorded through the distributed characteristics of the blockchain, the logic and algorithm implementation of the smart contract are continuously optimized according to the test results, relying on the characteristics of the smart contract, a small model based on the smart contract is established or combined with the incentive mechanism to promote the contribution and processing of compliant data, and a large model data supervision system is obtained in combination with the blockchain.

2. The method of claim 1, wherein, The data desensitization process includes data replacement, data generalization and data perturbation, and the desensitized data is obtained after data desensitization, and the desensitized data is encrypted by using SM9 national secret implicit signature, symmetric encryption algorithm or asymmetric encryption algorithm to realize data encryption protection, and finally encrypted large model data is obtained.

3. The method of claim 2, wherein, The data desensitization and encryption data are based on the non-tamperability and transparency of the blockchain, and in the whole process, the behaviors of data access, modification and processing are recorded to generate verifiable audit and supervision logs, in combination with the smart contract, the role-based access control RBAC or attribute-based access control ABAC is used to manage data access permissions to realize the data access control mechanism, wherein only authorized users or nodes can access the encrypted data; identity authentication and authorization are performed by using the digital identity on the blockchain, a blockchain-based identity authentication mechanism is designed, the large model training data is trained in the large model after privacy protection, and the trained data is sampled and analyzed in combination with the incentive mechanism, if the privacy is still exposed, the nodes and organizations participating in the audit are traced back according to the logs, and are rewarded and punished.

4. The method of claim 1, wherein, The preset consensus node uses a consensus mechanism to determine the authenticity and keyword bias of data in the large model. The accuracy of the algorithm in the blockchain system is verified through a designed experiment. The experiment is conducted by using a real data set or a synthetic data set to control the variable grouping. The data authenticity and bias detection in different scenarios are simulated. The experimental results are optimized. The performance of the system in terms of data authenticity and bias is analyzed according to the experimental results. The algorithm and system design are optimized accordingly. Positive feedback is formed to perform data compliance determination.

5. The method of claim 1, wherein, The smart contract and incentive mechanism algorithm are used to identify the training data for rewards and punishments. The data subject is encouraged to participate in data supervision and contribution. The data supervision party receives corresponding compensation. At the same time, low-quality data providers are eliminated through competition, forming a supervision and punishment mechanism.

6. The method of claim 1, wherein, The large model data supervision system includes an access layer, a data layer, a network layer, a consensus layer, a contract layer, and an application layer. The chain members in the supervision blockchain include large model data institutions and computer organization institutions. According to the business categories, the two parties perform corresponding functions as data induction parties and algorithm providers in the system. The access layer is a program calling layer. It is connected with the business interface of the alliance blockchain to execute corresponding algorithms for large model data and perform supervision. The data layer represents the classic chain data structure of the blockchain, which mainly includes Pre-Hash, timestamp, nonce, MerkleRoot, block height, and large model data supervision block. The distributed nodes in the network layer use encryption algorithm technology to implement identity authentication and access control for each participating node during encrypted data transmission. The consensus layer uses a benchmark consensus algorithm to determine the consensus root in the system to ensure that each node in the blockchain has information consistency. The contract layer integrates a small model. The automatically executed code in the contract layer can quickly respond to compliance detection and access permission requirements. Various smart contracts are used to implement conditional constraints in the large model data supervision process through specific scripts. The application layer is used to build an incentive mechanism.

7. A large model data supervision system based on a blockchain, which adopts the large model data supervision method based on a blockchain in any one of claims 1 to 6, characterized in that, The data encryption module is used to receive large model data, perform data desensitization and encryption operations on the large model data, and obtain encrypted large model data. The data verification module is used to input the encrypted large model data into a pre-established control variable grouping small model to perform privacy protection effectiveness verification, and obtain verified small model training data. The pre-established consensus node participates in the sampling verification process of the control variable grouping small model. The data incentive module is used to input the obtained verified small model training data into a large model for training, obtain training data, and identify the training data for rewards and punishments based on a smart contract and an incentive mechanism algorithm to obtain data supervision results. ​ 8. A terminal device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that, The memory stores a computer program capable of running on the processor, and when the processor loads and executes the computer program, a large model data supervision method based on the blockchain in any one of claims 1 to 6 is adopted.

Images