Large model corpus data interaction system and interaction method based on trusted execution environment, terminal and medium

Abstract

The invention discloses a large model corpus data interaction system and method based on a trusted execution environment, a terminal and a medium. The system comprises a data provider, a data demander and a trusted data space service platform. And the data provider is used for encrypting the corpus data, and performing preprocessing, directory sorting and packaging on the corpus data in the data provider connector to obtain the data capsule. And the data demander is used for submitting a calculation task in the data demander connector and obtaining corpus data of the data capsule for calculation after authorization. And the trusted data space service platform is used for storing the encrypted corpus data, controlling access of a data user, scheduling a calculation task, managing a key and generating an audit log. Through trusted execution environment hardware isolation and full-link encryption, it is ensured that corpus data only exists in a plaintext form in the trusted execution environment in the full life cycle from access to destruction, and unauthorized access and leakage are completely eradicated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of trusted execution environment, in particular to a large model corpus data interaction system and method based on trusted execution environment, a terminal and a medium. BACKGROUND

[0002] With the rapid development of generative AI and large model technology, the performance (such as understanding ability, generation quality) of large models is highly dependent on the support of high-quality and large-scale corpus data. At present, large model training parties (such as technology companies and research institutions) need to obtain corpus data from multiple channels (such as data providers and third-party data platforms), but there are the following core technical status and challenges in the corpus delivery process.

[0003] At present, data delivery usually adopts the following ways: 1. Direct clear text delivery: extremely high risk, easy to cause data leakage and misuse.

[0004] 2. Traditional encryption delivery: data needs to be decrypted before calculation, and the decrypted data still has the risk of being stolen in the memory (for example, through advanced malicious software or internal attacks of cloud platforms).

[0005] 3. Secure multi-party computation or homomorphic encryption: although these cryptographic schemes can guarantee the encryption of calculation, the calculation overhead is huge, the communication complexity is high, and it is difficult to apply to the processing scene of massive large model corpus data, and the practicality is limited.

[0006] Although there are some data sharing platforms based on software encryption and permission management in the prior art, these platforms provide certain access control, but their trust foundation is built on the software stack and operating system, which has a huge attack surface and cannot resist basic attacks from platform administrators or underlying systems.

[0007] The existing trusted technology is insufficient in security protection, specifically as follows: 1. Single TEE (Trusted Execution Environment) technology: although it can provide a hardware isolated trusted execution environment, the existing scheme focuses on the calculation trust of a single node, does not cover the whole link (such as access verification, transmission control, and use audit) from the data provider to the data demander, and is not optimized for the "massive and heterogeneous" characteristics of large model corpus data. Directly importing TB-level corpus data into TEE will cause memory overflow and high processing delay.

[0008] 2. Traditional trusted data space: existing trusted data space solutions focus on the trustworthiness of data storage and sharing, but their core mechanisms rely on software-level identity authentication and permission control, lacking hardware-level execution environment trust assurance. When data demanders decrypt corpus in the trusted data space and use it for large model training, they still face the risk of leakage caused by untrusted computing environment.

[0009] Therefore, the prior art still has defects. SUMMARY

[0010] The technical problem to be solved by the present application is to provide a large model corpus data interaction system and method based on a trusted execution environment, a terminal and a medium, which overcome the above-mentioned defects of the prior art. The technical solutions adopted by the present application are as follows: In a first aspect, the present application provides a large model corpus data interaction system based on a trusted execution environment, wherein the system comprises: a data provider configured to encrypt corpus data using an encryption algorithm and upload the corpus data to a trusted execution environment, and pre-process, catalog and package the corpus data in a data provider connector constructed in the trusted execution environment to obtain a data capsule; a data demander configured to submit a computing task in a data demander connector constructed in the trusted execution environment, and obtain corpus data of the data capsule for computing after obtaining authorization from the data provider; a trusted data space service platform configured to store encrypted corpus data, control access of data users, schedule computing tasks, manage keys and generate audit logs.

[0011] In an implementation manner, the system further comprises: a trusted execution environment cluster management platform configured to uniformly and cluster manage different types of trusted execution environments, including a trusted execution environment corresponding to the data provider and a trusted execution environment corresponding to the data demander.

[0012] In a second aspect, the present application also provides a large model corpus data interaction method based on a trusted execution environment, which is applied to the large model corpus data interaction system based on a trusted execution environment in the above-mentioned solution, and the method comprises: a data provider encrypting corpus data using an encryption algorithm and uploading the corpus data to a trusted execution environment, and pre-processing, cataloging and packaging the corpus data in a data provider connector constructed in the trusted execution environment to obtain a data capsule; a data demander initiating a data use request, forwarding the data use request to a data provider based on a trusted data space service platform, and obtaining corpus data in the data capsule for use after obtaining authorization from the data provider. monitoring the use behavior of the corpus data by the data demand side, and deleting the corpus data stored locally by the data demand side and clearing the corpus plaintext data in the trusted execution environment corresponding to the data demand side when the use period of the corpus data expires or the data demand side completes the use of the corpus data.

[0013] In an implementation manner, before the data provider encrypts the corpus data using an encryption algorithm and uploads to the trusted execution environment, further comprising: The trusted data space service platform deploys a public key infrastructure authentication center to issue digital certificates for the data provider, the data demand side, and the nodes of the trusted execution environment; The data provider registers and logs in the user identity through the data provider connector and registers and associates in the trusted data space service platform; The data demand side registers and logs in the user identity through the data demand side connector and registers and associates in the trusted data space service platform.

[0014] In an implementation manner, the data provider encrypts the corpus data using an encryption algorithm and uploads to the trusted execution environment, and pre-processes, organizes and packages the corpus data in the data provider connector constructed in the trusted execution environment to obtain a data capsule, comprising: The data provider deploys the trusted execution environment and deploys the corresponding trusted execution environment cluster management platform; The data provider encrypts the corpus data through the trusted key distributed by the data provider connector, and uploads the corpus data to the trusted execution environment of the data provider after remote verification of the trusted execution environment of the data provider is passed; The data provider connector decrypts the corpus data through the private key, and sets up a data directory and processes data for the corpus data in the data provider connector to form a corpus data product; The corpus data product is embedded in a digital watermark bound with the identity of the data provider, and is encrypted using the SM4 encryption algorithm to encapsulate as a data capsule; The data provider uploads the data capsule to the trusted data space service platform and performs on-chain.

[0015] In an implementation manner, after the data provider uploads the data capsule to the trusted data space service platform and performs on-chain, comprising: The data demand side deploys the trusted execution environment and deploys the corresponding trusted execution environment cluster management platform; The data demander builds a tool chain through the trusted execution environment, packs the model application, and uploads the model application to the trusted execution environment of the data demander after remote verification of the trusted execution environment of the data demander is passed; The data demander connector decrypts the model application through a private key; The data demander encrypts the model application and generates corresponding digest information, and uploads the model application to the trusted data space service platform for auditing.

[0016] In an implementation manner, the data demander initiates a data use request, and forwards the data use request to a data provider based on a trusted data space service platform, and obtains corpus data in a data capsule for use after authorization of the data provider, including: The data demander initiates a data use request, and the trusted data space service platform sets a data contract of corpus data requested by the data demander, and forwards the data use request to the data provider; The data provider authorizes the corpus data requested by the data demander, negotiates an encryption key through a data provider connector and a data demander connector, and transmits the data capsule to a trusted execution environment corresponding to the data provider; The data demander unpacks the data capsule in the corresponding trusted execution environment, and uses corpus data in the data capsule within an authorized range.

[0017] In an implementation manner, the use behavior of the data demander on the corpus data is monitored, and corpus data stored locally by the data demander is deleted, and corpus plaintext data in the trusted execution environment corresponding to the data demander is cleared after a use period of the corpus data expires or the data demander completes use of the corpus data, including: The use behavior of the data demander on the corpus data is obtained, and key data in a process in which the corpus data is used is chained and notarized; The use behavior of the corpus data is checked and audited, and when out-of-range use is found, authority recovery is triggered; Corpus data stored locally by the data demander is deleted, and corpus plaintext data in the trusted execution environment corresponding to the data demander is cleared after a use period of the corpus data expires or the data demander completes use of the corpus data; Data destruction logs are chained for remote verification by the data provider.

[0018] In a third aspect, the embodiments of the present application also provide a terminal, wherein the terminal comprises a memory, a processor, and a large model corpus data interaction program based on a trusted execution environment stored in the memory and executable on the processor, and the processor executes the large model corpus data interaction program based on the trusted execution environment to implement the steps of the large model corpus data interaction method based on the trusted execution environment in any of the above solutions.

[0019] In a fourth aspect, the embodiments of the present application also provide a computer readable storage medium, wherein the computer readable storage medium stores a large model corpus data interaction program based on a trusted execution environment, and the large model corpus data interaction program based on the trusted execution environment is executed by a processor to implement the steps of the large model corpus data interaction method based on the trusted execution environment in any of the above solutions.

[0020] Advantages: Compared with the prior art, the present application provides a large model corpus data interaction system based on a trusted execution environment, which comprises a data provider, a data demander, and a trusted data space service platform. The data provider is used to encrypt corpus data using an encryption algorithm and upload the corpus data to a trusted execution environment, and preprocess, catalog, and package the corpus data in a data provider connector constructed in the trusted execution environment to obtain a data capsule. The data demander is used to submit a computing task in a data demander connector constructed in the trusted execution environment, and obtain corpus data of the data capsule for computing after being authorized by the data provider. The trusted data space service platform is used to store the encrypted corpus data, control access of a data user, schedule a computing task, manage a key, and generate an audit log. The present application ensures that corpus data exists in clear text only in a trusted execution environment in the whole life cycle from access to destruction through hardware isolation and full-link encryption of the trusted execution environment, and prevents unauthorized access and leakage. In addition, the present embodiment integrates trusted data space and trusted execution environment technologies to cover the whole process from access verification, encrypted transmission, trusted processing, authorized delivery, use audit to safe destruction, and ensures traceability and non-tamperability of operations at each link. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The system architecture diagram of the large model corpus data interaction system based on the trusted execution environment provided by the embodiments of the present application in a distributed point-to-point delivery mode.

[0022] Figure 2 The system architecture diagram of the large model corpus data interaction system based on the trusted execution environment provided by the embodiments of the present application in a data and model centralized hosting mode.

[0023] Figure 3A flowchart of a preferred embodiment of the method for corpus data interaction of a large model based on a trusted execution environment provided by an embodiment of the present application.

[0024] Figure 4 A principle block diagram of a terminal provided by an embodiment of the present application. DETAILED DESCRIPTION

[0025] To make the objectives, technical solutions and effects of the present application clearer and more explicit, the present application is further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0026] The flowchart shown in the drawings is only an example and does not necessarily include all contents and operations or steps, nor does it necessarily be executed in the order described. For example, some operations or steps can be decomposed, combined or partially merged, so that the actual execution order can be changed according to the actual situation.

[0027] It should be understood that the terms used in the present application specification herein are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should be understood that, in order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the terms "first", "second" and the like are used to distinguish the same or similar items with basically the same function and effect. For example, the first control information and the second control information are only used to distinguish different control information, and do not limit the order. Those skilled in the art can understand that the terms "first", "second" and the like do not limit the quantity and execution order, and the terms "first", "second" and the like do not necessarily mean different. It should also be understood that the term "and / or" used in the present application specification and the appended claims means any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0028] In the existing corpus data interaction process, the following disadvantages exist: 1. High risk of privacy leakage: in the traditional encryption transmission and plaintext processing mode, the corpus data is exposed to the general computing environment after decryption, and the internal personnel of the data demand party and malicious software can obtain the corpus plaintext, and there is a lack of effective leakage tracing means.

[0029] 2. TEE (Trusted Execution Environment) and large model corpus adaptability is poor: existing TEE solutions are not optimized for massive corpus, and directly processing TB-level corpus will cause insufficient memory and low processing efficiency. For example, according to an experiment, it takes 72 hours to process 1 TB of text corpus under the traditional scheme, which is far beyond the time requirement of large model training, and cannot be compatible with the unified processing of heterogeneous corpus such as text, image and voice.

[0030] 3. Full-process credibility is broken: existing trusted data space only covers storage and sharing links, and TEE only covers computing links, and the two are not coordinated. After the corpus completes identity verification in the trusted data space, there is a lack of trusted connection in the process of transferring to TEE, and there is no subsequent audit mechanism when the processing result (such as pre-processed corpus) in TEE is delivered to the data demander, resulting in the problem of "trusted access and processing, but untrusted transmission or delivery".

[0031] 4. There is no basis for right confirmation and income distribution: the existing scheme lacks corpus data ownership identification and use tracking technology, and the data provider cannot prove the ownership of the corpus data, nor can it track the actual use range of the data demander (such as whether the corpus data is authorized to a third party), resulting in the problem of difficult right confirmation and difficult distribution.

[0032] 5. Performance and security imbalance: the performance loss of high security cryptographic scheme is too large, which cannot meet the requirement of high efficiency of large model data processing.

[0033] 6. Poor cross-platform compatibility: different TEE architectures and trusted data space protocols are not compatible, resulting in repeated adaptation for cross-enterprise corpus delivery, high research and development cost and difficulty in large-scale.

[0034] In order to solve the problems of the prior art, the present application provides a large model corpus data interaction system based on a trusted execution environment, specifically as follows Figure 1 and Figure 2As shown in the figure, the system includes a data provider, a data demander and a trusted data service space platform. Specifically, the data provider is the original owner of the corpus data, used to encrypt the corpus data using an encryption algorithm and upload to the trusted execution environment, and preprocess, catalog and package the corpus data in the data provider connector built in the trusted execution environment to obtain data capsules. The data demander is used to submit a computing task in the data demander connector built in the trusted execution environment, and after obtaining the authorization of the data provider, obtain the corpus data of the data capsule for calculation. The trusted data space service platform is used to store the encrypted corpus data, control the access of the data user, schedule the computing task, manage the key and generate the audit log. In addition, the system of the embodiment also includes a trusted execution environment cluster management platform for unified cluster management of different types of trusted execution environments, including a plurality of TEE cluster nodes, which in the embodiment can specifically include the trusted execution environment corresponding to the data provider and the trusted execution environment corresponding to the data demander.

[0035] In actual application, the trusted data space service platform of the embodiment is deployed on the trusted execution environment cluster management platform, and the trusted data space service platform includes three logical units, namely a non-secure area, an access control and task scheduling unit and a TEE secure area. The non-secure area is used to store the encrypted massive corpus data. The access control and task scheduling unit runs outside the TEE and is used to receive and process the request of the data demander, verify the identity and authority of the data demander, and schedule the authorized computing task to the TEE for execution. The TEE secure area is the clustered trusted execution environment, which runs a trusted data processing program inside, and the program includes a key processing module, a secure computing engine and an audit log generation module, etc. The trusted execution environment cluster management platform includes various different TEE schemes and performs pool management on the TEE, supporting users to deploy applications through kubernetes, etc. Kubernetes is an open source application for managing containerized applications on multiple hosts in a cloud platform.

[0036] In combination with Figure 1 and 2As shown in the foregoing embodiments, the data provider connector and the data demander connector are both deployed in the trusted execution environment, and the trusted data space service platform is also deployed in the trusted execution environment, thus forming a two-layer architecture of the trusted data space management layer and the trusted execution layer, which is conducive to realizing the trusted closed-loop management and control of the whole process of the large model corpus data from access to transmission to processing to delivery to destruction, and solving the core problems of the untrusted environment, uncontrollable process and easy leakage of privacy in the traditional corpus delivery. In addition, in the present embodiment, a data capsule is used as the core carrier for corpus data delivery. The data capsule is an integrated encapsulation body that is generated in the trusted execution environment and deeply binds the encrypted corpus data and the digital contract, aiming to realize the control right of corpus data delivery. The core components of the data capsule of the present embodiment include a core layer, a control layer, a right confirmation layer and a metadata layer. The core layer is large model corpus encrypted by the SM4 algorithm, specifically including heterogeneous data such as text, image and voice, and these large model corpus can only be decrypted in the authorized TEE. The control layer includes a digital contract, which contains initial use rules such as corpus use range, number of times, time length, destruction condition, income distribution ratio, etc., and the data contract supports subsequent dynamic negotiation and adjustment. The right confirmation layer includes invisible digital watermark bound with the identity of the data provider and corpus digest signature based on the SM3 algorithm (a commercial cryptographic hash algorithm), the invisible digital watermark is used for tamper-proofing, and the corpus digest signature is used for anti-counterfeiting. The metadata layer includes corpus type, size, right information, generation time, TEE node identifier, etc., which are used to support platform directory management and blockchain traceability. The data capsule can be generated, decrypted and executed in the TEE throughout the process, preventing unauthorized access, and the circulation records (such as generation, authorization, delivery and destruction) of the data capsule can be stored on the chain for evidence, ensuring traceability throughout the life cycle.

[0037] Based on the foregoing embodiments, the present application also provides a large model corpus data interaction method based on a trusted execution environment, which can be applied to the system of the present embodiment. In actual application, the method of the present embodiment can be applied to a terminal, which can be a computer or other intelligent product terminal, and the terminal can be used to execute the large model corpus data interaction method based on the trusted execution environment. Specifically, as shown in the foregoing embodiments, the large model corpus data interaction method based on the trusted execution environment of the present embodiment includes the following steps: Figure 3 Step S100, the data provider uses an encryption algorithm to encrypt the corpus data and upload it to the trusted execution environment, and pre-processes, organizes and packages the corpus data in the data provider connector constructed in the trusted execution environment to obtain a data capsule.

[0038] ​Firstly, the trusted data space service platform deploys a public key infrastructure authentication center to issue digital certificates for data providers, data demanders, and nodes of the trusted execution environment. Next, the data provider registers and logs in the user identity through the data provider connector and registers and associates in the trusted data space service platform. Then, the data demander registers and logs in the user identity through the data demander connector and registers and associates in the trusted data space service platform.

[0039] Further, the data provider deploys the trusted execution environment and deploys the corresponding trusted execution environment cluster management platform. Then, the data provider encrypts the corpus data through the trusted key pair issued by the data provider connector, and uploads the corpus data into the trusted execution environment of the data provider after passing the remote verification of the trusted execution environment of the data provider. Next, the data provider connector decrypts the corpus data through the private key, at this time only the TEE can decrypt the corpus data, and malicious personnel or high-privilege administrators cannot decrypt the data. The data provider sets up a data directory and processes the corpus data in the data provider connector to form a corpus data product. Then, the corpus data product is embedded in the digital watermark bound with the identity of the data provider, and is encrypted by using the SM4 encryption algorithm (a symmetric encryption algorithm) to encapsulate as a data capsule, at this time the data capsule includes encrypted corpus, digital contract, digital watermark, and corpus digest signature based on the SM3 algorithm. Finally, the data provider uploads the data capsule to the trusted data space service platform and performs on-chain.

[0040] Further, the data demander deploys the trusted execution environment and deploys the corresponding trusted execution environment cluster management platform. Next, the data demander packs the model application through the trusted application construction tool chain of the trusted execution environment, and uploads the model application into the trusted execution environment of the data demander after passing the remote verification of the trusted execution environment of the data demander. Then, the data demander connector decrypts the model application through the private key, at this time only the TEE can decrypt the user data, and malicious personnel or high-privilege administrators cannot decrypt the data. The data demander encrypts the model application and generates corresponding digest information, uploads to the trusted data space service platform for audit, and only the model application after passing the audit can be used.

[0041] Step S200, the data demander initiates a data use request, and forwards the data use request to the data provider based on the trusted data space service platform, and after obtaining the authorization of the data provider, obtains the corpus data in the data capsule for use.

[0042] The data demander discovers corpus data products in the data space service platform, initiates a data use request, and the trusted data space service platform sets a data contract for the corpus data requested by the data demander and forwards the data use request to the data provider. The data contract includes: limiting data product content, limiting data product delivery connector, limiting data product use connector, limiting data product user, limiting data product use operation, limiting data product use times, limiting data product use time length, limiting data product visible range, limiting data product available range, data product use time notification, data product use after destruction, etc. Then, the data provider authorizes the corpus data requested by the data demander, negotiates an encryption key through the data provider connector and the data demander connector, and transmits the data capsule to the trusted execution environment corresponding to the data provider. After delivery is completed, the data demander unpacks the data capsule in the trusted execution environment and uses the corpus data in the data capsule within the authorized range.

[0043] Step S300, monitoring the use behavior of the data demander to the corpus data, and deleting the corpus data stored locally by the data demander and clearing the corpus plaintext data in the trusted execution environment corresponding to the data demander after the corpus data use period expires or the data demander completes the use of the corpus data.

[0044] The embodiment based on the data capsule constructs an access control strategy based on a smart contract in the trusted data space, checks and audits the use behavior of the corpus data. Specifically, the embodiment obtains the use behavior of the data demander to the corpus data, chains and notarizes the key data in the process of using the corpus data. Then, the use behavior of the corpus data is checked and audited, and when the use is found to be out of range, the authority is recovered, that is, the TEE node stops supplying the corpus data. After the corpus data use period expires or the data demander completes the use of the corpus data, the embodiment can trigger the deletion of the corpus data stored locally by the data demander and the clearing of the corpus plaintext data in the trusted execution environment corresponding to the data demander, and chain the data destruction log, and the data demander can provide the data destruction log to the data provider for remote verification and update the data contract.

[0045] Therefore, this embodiment proposes a two-layer architecture of a trusted data space management layer and a trusted execution layer. This facilitates a trusted closed-loop management of large-scale model corpus data throughout the entire process from access to transmission, processing, delivery, and destruction, solving the core problems of untrustworthy environments, uncontrollable processes, and easy privacy leaks in traditional corpus delivery. Furthermore, this embodiment combines digital watermarking with blockchain notarization, completing corpus watermark generation and verification within the TEE. Simultaneously, it stores corpus data delivery records (such as data provider, data requester, scope of corpus data use, and validity period of corpus data use) on the blockchain, achieving corpus data ownership confirmation and usage traceability, supporting subsequent revenue distribution. In addition, this embodiment constructs a smart contract-based access control policy within the trusted data space based on a data capsule approach, and can integrate data sandbox technology. Combined with the TEE's hardware-level identity authentication, it enables dynamic adjustment of corpus data usage permissions (such as temporary authorization and recycling after use), preventing overuse.

[0046] In practical use, the trusted data space service platform of this embodiment supports a distributed peer-to-peer delivery mode, as well as a centralized data and model hosting mode. In the distributed peer-to-peer delivery mode, such as... Figure 1 As shown. The specific process is as follows: The data provider performs step 1-1. Identity registration / authentication, and the data requester performs step 1-2. Identity registration / authentication. Next, the data provider performs step 2-1. Encrypted upload of corpus data, and the data requester performs step 2-2. Encrypted upload of model application. Further, in the trusted execution environment, the data provider connector performs step 3-1. Reporting the data product catalog and uploading it to the trusted data space service platform. In the trusted execution environment, the data provider connector performs step 3-2. Reporting model information for review. Next, the data provider connector performs step 4. Initiating a corpus data application. The trusted data space service platform performs step 5. Issuing a data application request and negotiating a digital contract. The data requester connector performs step 6. Returning the review result, and the trusted data space service center performs step 7. Returning the approval result. Then, the data provider connector and the data requester connector perform step 8. Negotiating a key to build an encrypted channel. The data provider performs step 9. Delivering the data capsule. The data requester connector and the data provider connector perform step 10. Delivery and usage monitoring, and completing the data contract update. Therefore, in the distributed peer-to-peer interaction mode, the data demander connector and the data provider connector can deliver data capsules to each other, achieving peer-to-peer encrypted delivery.

[0047] The trusted execution environment enables functions such as identity management, access connector management, directory management, trusted data space management, digital contract management, data usage control, key management, model hosting, and data hosting. Therefore, the trusted data space service platform in this embodiment also supports a centralized data and model hosting mode. In the centralized data and model hosting mode, such as... Figure 2 As shown, the specific process is as follows: The data provider performs step 1-1. Identity registration / authentication, and the data requester performs step 1-2. Identity registration / authentication. Next, the data provider performs step 2-1. Encrypted upload of the corpus data, and the data requester performs step 2-2. Encrypted upload of the model application. Further, in the trusted execution environment, the data provider connector performs step 3-1. Reporting the data product catalog and uploading it to the trusted data space service platform. In the trusted execution environment, the data provider connector also performs step 3-2. Reporting the encrypted model application. Next, the data provider connector performs step 4. Uploading the data capsule containing the encrypted corpus data. The data requester connector performs step 5. Viewing the corpus data catalog and step 6. Initiating a corpus data request. Then, the data provider connector and the data requester connector perform step 7. Digital contract negotiation, followed by the data provider performing step 8. Reviewing the digital contract negotiation. Finally, the data requester connector and the trusted data space service platform perform step 9. Initiating the delivery of the data capsule containing the corpus data for model training, fine-tuning, and inference. This demonstrates that, under the centralized hosting model for data and models, the delivery of corpus data is accomplished by the trusted data space service platform.

[0048] Compared with the prior art, the present invention has the following significant advantages: 1. Enhanced security: Utilizing the hardware-level security isolation of TEE, sensitive data is ensured to remain encrypted even during computation, effectively defending against attacks from the operating system and even physical hardware.

[0049] 2. Higher performance: Plaintext data is processed directly within the TEE, resulting in extremely high computational efficiency and minimal performance loss, making it ideal for processing massive large-scale model corpora.

[0050] 3. Verifiable trust: By recording access policies and audit logs through blockchain, the data usage process is made transparent and tamper-proof, establishing a trust foundation for multi-party collaboration.

[0051] 4. End-to-end protection: It achieves full-process security protection from data delivery and computation to result return, covering the entire data lifecycle.

[0052] The various modules in the above large model corpus data interaction system based on the trusted execution environment can be realized by software, hardware, and combinations thereof, in whole or in part. The various modules can be embedded in or independent of a processor in the terminal in hardware form, or can be stored in a memory in the terminal in software form, so as to be called and executed by the processor to perform the operations corresponding to the various modules.

[0053] Based on the above embodiments, the present application further provides a terminal, a principle block diagram of which can be as shown in Figure 4 The terminal can include one or more processors 100 (only one is shown in Figure 4 The one or more processors 100 can implement the various steps in the sleep analysis method based on multi-sensor data embodiments when executing the computer program 102. Alternatively, the one or more processors 100 can implement the functions of the modules / units in the large model corpus data interaction system based on the trusted execution environment embodiments when executing the computer program 102, which is not limited here.

[0054] In an embodiment, the processor 100 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0055] In an embodiment, the memory 101 can be an internal storage unit of the electronic device, such as a hard disk or a memory of the electronic device. The memory 101 can also be an external storage device of the electronic device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the memory 101 can include both the internal storage unit and the external storage device of the electronic device. The memory 101 is used to store computer programs and other programs and data required by the terminal. The memory 101 can also be used to temporarily store data that has been output or will be output.

[0056] Those skilled in the art can understand that, Figure 4 The principle block diagram shown in the above-mentioned embodiments is only a block diagram of part of the structure related to the technical scheme of the present application, and does not constitute a limitation on the terminal to which the technical scheme of the present application is applied. The specific terminal can include more or less components than those shown in the diagram, or combine certain components, or have a different arrangement of components.

[0057] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing relevant hardware. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiments. Any reference to memory, storage, operating database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM) and the like.

[0058] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A trusted execution environment-based large model corpus data interaction system, characterized in that, The system comprises: a data provider for encrypting corpus data using an encryption algorithm and uploading to a trusted execution environment, and preprocessing, cataloging and packaging the corpus data in a data provider connector built in the trusted execution environment to obtain a data capsule; a data demander for submitting a computing task in a data demander connector built in the trusted execution environment, and obtaining corpus data of the data capsule for computing after authorization by the data provider; a trusted data space service platform for storing the encrypted corpus data, controlling access of data users, scheduling computing tasks, managing keys and generating audit logs.

2. The trusted execution environment-based large model corpus data interaction system according to claim 1, wherein, The system further comprises: a trusted execution environment cluster management platform for unified cluster management of different types of trusted execution environments, including the trusted execution environment corresponding to the data provider and the trusted execution environment corresponding to the data demander.

3. A method for data interaction of a large model corpus based on a trusted execution environment, the method being applied to the system for data interaction of a large model corpus based on a trusted execution environment according to claim 1 or 2, characterized in that, The method comprises: the data provider encrypting corpus data using an encryption algorithm and uploading to a trusted execution environment, and preprocessing, cataloging and packaging the corpus data in a data provider connector built in the trusted execution environment to obtain a data capsule; the data demander initiating a data use request, and forwarding the data use request to the data provider based on the trusted data space service platform, and obtaining corpus data in the data capsule for use after authorization by the data provider; monitoring the use behavior of the data demander on the corpus data, and deleting the corpus data stored locally by the data demander and clearing the corpus plaintext data in the trusted execution environment corresponding to the data demander when the corpus data use period expires or the data demander completes the use of the corpus data.

4. The trusted execution environment-based large model corpus data interaction method according to claim 3, characterized in that, Before the data provider encrypts corpus data using an encryption algorithm and uploads to a trusted execution environment, it further comprises: the trusted data space service platform deploying a public key infrastructure authentication center to issue digital certificates for the data provider, the data demander and the nodes of the trusted execution environment; the data provider registering and logging in the user identity through the data provider connector, and registering and associating in the trusted data space service platform; the data demander registering and logging in the user identity through the data demander connector, and registering and associating in the trusted data space service platform.

5. The trusted execution environment-based large model corpus data interaction method according to claim 3, characterized in that, The data provider encrypts corpus data using an encryption algorithm and uploads to a trusted execution environment, and preprocesses, catalogs and packages the corpus data in a data provider connector built in the trusted execution environment to obtain a data capsule, comprising: the data provider deploying a trusted execution environment and deploying a corresponding trusted execution environment cluster management platform; the data provider encrypting corpus data through a trusted key distributed by the data provider connector, and uploading the corpus data to the trusted execution environment of the data provider after remote verification of the trusted execution environment of the data provider is passed; the data provider connector decrypting the corpus data through a private key, and setting a data catalog and processing data of the corpus data in the data provider connector to form a corpus data product; the data provider connector decrypting the corpus data through a private key, and setting a data catalog and processing data of the corpus data in the data provider connector to form a corpus data product; Embed the corpus data product in a digital watermark bound to the data provider identity, and encrypt the corpus data product using the SM4 encryption algorithm, encapsulating it as a data capsule; The data provider uploads the data capsule to the trusted data space service platform and performs on-chain.

6. The trusted execution environment-based large model corpus data interaction method according to claim 5, characterized in that, After the data provider uploads the data capsule to the trusted data space service platform and performs on-chain, the following steps are included: The data demander deploys a trusted execution environment and deploys a corresponding trusted execution environment cluster management platform; The data demander constructs a tool chain through the trusted execution environment, packages the model application, and after remote verification of the data demander's trusted execution environment is passed, uploads the model application to the data demander's trusted execution environment; The data demander connector decrypts the model application through a private key; The data demander encrypts the model application and generates corresponding digest information, and uploads it to the trusted data space service platform for audit.

7. The trusted execution environment-based large model corpus data interaction method according to claim 3, characterized in that, The data demander initiates a data use request, and based on the trusted data space service platform, forwards the data use request to the data provider, and after obtaining the authorization of the data provider, obtains the corpus data in the data capsule for use, including: The data demander initiates a data use request, and the trusted data space service platform sets a data contract for the corpus data requested by the data demander, and forwards the data use request to the data provider; The data provider authorizes the corpus data requested by the data demander, negotiates an encryption key through the data provider connector and the data demander connector, and transmits the data capsule to the data provider's corresponding trusted execution environment; The data demander unpacks the data capsule in the corresponding trusted execution environment and uses the corpus data in the data capsule within the authorized scope.

8. The trusted execution environment-based large model corpus data interaction method according to claim 3, characterized in that, Monitor the data demander's use of the corpus data, and after the corpus data usage period expires or the data demander completes the use of the corpus data, delete the corpus data stored locally by the data demander, and clear the corpus plaintext data in the data demander's corresponding trusted execution environment, including: Obtain the data demander's use of the corpus data, chain and notarize the key data in the process of using the corpus data; Verify and audit the use of the corpus data, and trigger authority recovery when out-of-scope use is found; After the corpus data usage period expires or the data demander completes the use of the corpus data, delete the corpus data stored locally by the data demander, and clear the corpus plaintext data in the data demander's corresponding trusted execution environment; Chain the data destruction log for remote verification by the data provider.

9. A terminal, characterized by comprising: The terminal includes a memory, a processor, and a large model corpus data interaction program based on a trusted execution environment stored in the memory and executable on the processor. When the processor executes the large model corpus data interaction program based on the trusted execution environment, it implements the steps of the large model corpus data interaction method based on the trusted execution environment as claimed in any one of claims 3-8.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a trusted execution environment-based large model corpus data interaction program. When the trusted execution environment-based large model corpus data interaction program is executed by the processor, the steps of the trusted execution environment-based large model corpus data interaction method according to any one of claims 3-8 are implemented.

Images