A hardware root-of-trust-based end-side idle time computing power sharing and secure transaction method and system

Abstract

The application discloses an end-side idle time computing power sharing and secure transaction method and system based on a hardware root of trust, and belongs to the technical field of distributed computing and data security. A hardware-level logically isolated secure execution environment is created in a trusted execution environment through a hardware root of trust built in an end-side device, so that bidirectional isolation of task data and device data is ensured. Each transaction is jointly signed by the hardware roots of trust of both parties, so that an unalterable transaction settlement voucher is formed. The system supports user control functions such as manually accepting or rejecting a request, starting or stopping idle time sharing, and externally requesting computing power. After the execution environment is destroyed, full zero overwrite is performed, and a verifiable data erasure certificate is generated. The application solves the core problem of bidirectional trust in idle time computing power sharing, supports transaction evidence storage and public welfare donation evidence storage in accordance with legal norms, and provides a secure and reliable technical foundation for constructing a global end-side distributed computing power network.

Description

Technical Field

[0001] This invention relates to the fields of distributed computing and data security technology, specifically to a method and system for secure sharing and trading of idle computing resources of edge computing devices. Background Technology

[0002] With the widespread adoption of AI technology, billions of edge computing devices worldwide remain idle most of the time. The total idle computing power of these devices far exceeds that of any single supercomputing center. However, current technologies cannot effectively utilize these idle computing resources while ensuring the data security of computing power providers. Existing distributed computing solutions suffer from the following drawbacks: computing power providers cannot be certain that their devices will not be attacked by malicious code; computing power users cannot be certain that their computing task data will not be stolen by computing power providers; and there is a lack of a reliable transaction settlement mechanism, preventing the formation of an effective computing power trading market. Existing distributed computing solutions are mostly designed for cloud servers and dedicated edge nodes, and do not address the unique pain points of personal end-device devices, such as the uncertainty of idle time, user control needs, and the bottom line of device hardware security, to build a complete secure transaction closed loop. Summary of the Invention

[0003] This invention aims to solve the security and trust issues in sharing idle computing power on the edge. By establishing a secure execution environment logically isolated from the main operating system at the hardware level within a trusted execution environment through a hardware root of trust, this secure execution environment includes a secure computing power container or a hardware-mandated execution environment, ensuring that the computing power user's task data cannot be accessed by the computing power provider; simultaneously, it ensures that the computing power provider's equipment cannot be attacked by malicious task code. Each transaction is jointly signed by the hardware roots of trust of both parties, forming an immutable transaction settlement certificate. Transaction records jointly signed by the hardware roots of trust of both parties meet the requirements of reliable electronic signatures stipulated in the Electronic Signature Law, possessing legal evidentiary effect and financial document attributes. The distributed computing power scheduling layer described in this article refers to the logical functional layer that realizes the registration and authentication of computing power nodes, the matching and scheduling of computing power resources, and the transaction settlement and storage. It can be implemented in any form, such as a centralized cloud platform, a distributed node cluster, a P2P scheduling protocol, or an end-side scheduling module, and is not limited by the specific deployment location or operating entity. The core improvement of this invention lies in applying hardware root of trust technology to the idle computing power sharing scenario on the edge. Through a secure execution environment, a secure closed loop is achieved, ensuring "data does not leave the environment and computing power is destroyed after use," thus solving the long-standing two-way trust problem in idle computing power sharing. Simultaneously, it provides user control functions such as manually accepting or rejecting, manually enabling or disabling idle sharing, manually setting idle time periods, and automatically or manually requesting computing power from external sources, giving users complete control over computing power sharing. The distributed computing power scheduling layer can also automatically complete donation transfers and generate tamper-proof donation records based on the donation rules preset by the computing power provider. Attached Figure Description

[0004] Figure 1 This is a diagram of the overall system architecture of the present invention; Figure 2 This is a flowchart of the process for sharing computing power during off-peak hours and securing transactions. Detailed Implementation

[0005] The present invention will now be described in detail with reference to the accompanying drawings and embodiments: Example 1: Sharing Idle Computing Power of Personal AI Hosts User A purchased an edge AI computing device for daily work and home AI services. During the early morning hours, the device is in idle mode. User A enabled the "Idle Computing Power Sharing" function in the device settings and set idle time criteria: a basic condition of no device interaction for more than 30 minutes, plus additional conditions that the device is charging and the device temperature is below 60 degrees Celsius. User A also set a minimum acceptable unit price and a maximum daily sharing time to ensure that computing power sharing would not affect the device's lifespan. User A also set a donation rule: the first 100 yuan of revenue each month will be automatically donated to a children's education charity. Once the idle time conditions are met, the device's hardware root of trust automatically measures and signs the currently available computing power, generates an idle time computing power listing record, and publishes it to the distributed computing power scheduling layer. The listing price is 0.01 yuan per computing power unit. At this time, an AI model company needs a large amount of computing power for model fine-tuning. The company initiates a computing power usage request through a distributed computing power scheduling layer, requiring 100 computing power units for inference. Based on the request, the distributed computing power scheduling layer matches 100 idle and trusted computing power nodes, including user A. After a successful match, User A's device hardware root of trust creates a secure computing container within the Trusted Execution Environment. This container is directly managed by the hardware root of trust and is logically isolated from the main operating system at the hardware level. Even if the highest administrator privileges on the main operating system are obtained, no data within the container can be accessed. The container is only granted the minimum CPU and memory permissions necessary to complete this fine-tuning task. In another embodiment, the secure execution environment can be implemented in the form of a hardware-isolated virtual machine built into the processor, a physical CPU core partition, etc., and its life cycle, resource permissions and data access boundaries are uniformly managed by the hardware root of trust to achieve the same hardware-level isolation effect. The AI ​​model company sends encrypted fine-tuning task data and runtime environment images to user A's device. This image, a standard template provided by the distributed computing power scheduling layer, is loaded into a secure computing power container after its integrity is verified by the hardware root of trust. During task execution, the hardware root of trust monitors the code behavior within the container in real time to prevent malicious code, unauthorized access, or mining and other violations. Upon completion of the task, the hardware root of trust signs the computation result and returns it. The AI ​​model company's hardware root of trust verifies the signature's validity and confirms the transaction's completion. The transaction record jointly signed by both parties' hardware roots of trust is stored in tamper-proof storage media. After the secure computing container is destroyed, the hardware root of trust performs a full-zero overwrite on the memory and storage area occupied by the container and generates a data erasure certificate with a hardware signature, allowing both parties to verify offline that no data remains. The following morning, User A checked the earnings report and found that the off-peak computing power sharing during the early morning hours had earned him 12.5 yuan. The first 100 yuan of this earnings had been automatically donated by the distributed computing power scheduling layer during transaction settlement, according to User A's preset donation rules. An immutable donation record signed by the hardware trust root was generated, which can serve as evidence for tax declarations and social responsibility reports. User A gave the service a five-star rating, which, after being signed by the hardware trust root, updated his credit score in the distributed computing power scheduling layer. The settlement record, after being jointly signed by both parties' hardware trust roots, generated a certificate of authenticity conforming to legal electronic transaction voucher standards, for use in tax declarations and financial audits. Example 2: The computing power user manually requests computing power from outside. During a major e-commerce promotion, User B needed to handle a large number of customer inquiries, but their local device lacked sufficient computing power. User B manually initiated an external computing power request through the user interface, inputting the task type, estimated computing power consumption, and maximum acceptable price. Based on User B's input specifications, the system automatically matched and recommended three optimal computing power nodes, displaying each node's credit score and estimated cost. User B manually selected the node with the highest score and clicked confirm. The system automatically completed the creation of a secure execution environment, task transfer, and result retrieval. Upon task completion, the system automatically settled the fee and deducted it from User B's account. Example 3: Manually accepting or rejecting computing power requests One day, User A received a request to use computing power. User A viewed the request details through the user interface, including the task type as "data analysis," the estimated power consumption of 20 units, the estimated revenue of 0.2 yuan, and the computing power user's credit score of 4.5. Because the user's credit score was lower than User A's preset automatic acceptance threshold of 4.8, the system prompted User A to make a manual decision. After reviewing the user's historical reviews, User A selected to accept. The system then proceeded to the secure execution environment creation and execution process; Example 4: Compensation processing for abnormal interruptions During the execution of the aforementioned task, User A needs to temporarily wake up the device due to an urgent matter. Upon detecting the user interaction, the system immediately triggers the task interruption protection mechanism. Unfinished tasks are automatically migrated to other idle computing power nodes for continued execution. The additional scheduling costs incurred due to the interruption are borne by the distributed computing power scheduling layer, and User A still receives proportional revenue for the completed computing power portion. Responsibility determination is based on the task execution logs and interruption records signed by both parties' hardware trust roots. The logs are immutable, and the determination results are traceable and verifiable. Example 5: Detection and Blocking of Malicious Tasks A malicious user attempted to submit a task containing mining code through the distributed computing power scheduling layer. After the task data was sent to a certain edge device, the hardware root of trust detected the continuous high-load mining violation in real time while executing the task within the secure computing power container. The system immediately terminated the current task, forcibly destroyed the secure computing power container, and recorded the violation in the malicious user's credit file, permanently restricting their future access to the distributed computing power scheduling layer.

Claims

1. A method for sharing idle computing power and secure transactions on the edge based on a hardware root of trust, applied to a distributed computing power sharing network composed of multiple edge computing devices, characterized in that, Includes the following steps: Computing power registration and security authentication steps: The computing power provider generates a unique computing power node identity for the device through the hardware root of trust built into its edge computing device, and submits the device hardware configuration, available computing power specifications, idle time period and security level declaration to the distributed computing power scheduling layer after digitally signing the digital signature; the distributed computing power scheduling layer verifies the digital signature, and registers the device as a trusted computing power node after the verification is successful; Idle State Determination and Resource Publication Steps: Based on preset idle state determination conditions, the edge computing device automatically measures and signs the currently available computing power resources through the hardware root of trust after detecting that the device has entered an idle state, and generates an idle computing power listing record, which is then published to the distributed computing power scheduling layer; Computing Power Transaction Matching and Secure Execution Environment Creation Steps: When the edge device of the computing power user detects that its own computing power is insufficient or the user actively initiates an external computing power request, the distributed computing power scheduling layer matches one or more trusted computing power nodes that meet the conditions according to the computing power requirement specifications, security level requirements, and price budget in the request; Upon successful matching, on the edge computing device of the selected computing power provider, a secure execution environment is created within the trusted execution environment by the hardware root of trust. This secure execution environment is directly managed by the hardware root of trust and is logically isolated from the main operating system at the hardware level. The secure execution environment can be a secure computing power container or a hardware-enforced isolated execution environment. Even if the highest administrator privileges of the main operating system are obtained, no data, code, or runtime status within this secure execution environment can be accessed. The secure execution environment is only granted the minimum permissions necessary to complete the current computing task and cannot access any personal data of the computing power provider. Encrypted task transmission and trusted execution steps: The computing power user sends the task data to be computed and the runtime environment image to the matched computing power provider through an end-to-end encrypted channel. The hardware root of trust of the computing power provider performs integrity verification on the received task data and runtime environment image and executes the computing task within the secure execution environment. During task execution, the hardware root of trust monitors code behavior within the secure execution environment in real time. Throughout the entire computation process, neither the computing power provider's main operating system nor the user can access the task data, intermediate results, and running status within this secure execution environment; Result verification and transaction settlement steps: After the calculation task is completed, the hardware trust root digitally signs the calculation result and returns the signed result to the computing power user through an encrypted channel; after the computing power user's hardware trust root verifies the validity of the result signature, it confirms that the transaction is completed; after the transaction record is jointly signed by the hardware trust roots of both parties, it is stored in a tamper-proof storage medium as an immutable transaction settlement certificate.

2. The method according to claim 1, characterized in that, The lifecycle of the secure execution environment is bound to the current computing task; after the task is completed, the hardware root of trust forcibly reclaims all resources of the secure execution environment and erases all data within the secure execution environment to ensure that no task data of the computing power user remains on the computing power provider's device.

3. The method according to claim 1, characterized in that, When matching computing power nodes, the distributed computing power scheduling layer comprehensively considers the node's historical task completion rate, response latency, security level, and user evaluation. It calculates the adaptation score of each candidate node through a multi-attribute decision model and prioritizes the node with the highest score.

4. The method according to claim 3, characterized in that, After each computing power transaction is completed, the computing power user can evaluate and sign the service through its hardware root of trust. The evaluation record is stored in the distributed computing power scheduling layer. The computing power provider's credit score is dynamically updated based on historical evaluation records. The credit score affects its subsequent matching priority and transaction pricing.

5. The method according to claim 1, characterized in that, The idle time determination conditions include the basic determination condition that the device has no user interaction time for more than a preset threshold, and also include at least one additional determination condition that the device is in a charging state and the device temperature is within a safe range.

6. The method according to claim 1, characterized in that, The computing power provider can set a minimum price, charge by computing power unit or charge by task pricing strategy; the distributed computing power scheduling layer supports multi-party price comparison and automatic price negotiation functions.

7. The method according to claim 1, characterized in that, The runtime environment image is provided by the computing power user or by a standard image template that has undergone security auditing provided by the distributed computing power scheduling layer. Before the image is loaded into the secure execution environment, the integrity of the image is verified by the hardware root of trust. The image can only be loaded and run after the verification is passed.

8. The method according to claim 1, characterized in that, During task execution, the hardware root of trust monitors code behavior within the secure execution environment in real time. When any violation is detected, such as malicious code, unauthorized access to hardware resources, abnormal external packet sending, or continuous high-load mining, the current task is immediately terminated, the secure execution environment is forcibly destroyed, and the violation is recorded in the credit file of the computing power user, restricting their subsequent usage rights.

9. The method according to claim 2, characterized in that, After the secure execution environment is destroyed, the hardware trust root performs a full zero overwrite on the memory and storage areas occupied by the secure execution environment and generates a data erasure certificate with a hardware signature; the data erasure certificate can be verified offline by the computing power provider and the computing power user to confirm that there is no residual data in this task.

10. The method according to claim 1, characterized in that, The computing power provider can manually enable or disable the off-peak computing power sharing function through the user interface; when the function is disabled, even if the device meets the off-peak judgment conditions, the off-peak computing power listing record will not be published; when the function is enabled, the system will automatically trigger it according to the off-peak judgment conditions set by the user.

11. The method according to claim 1, characterized in that, When the distributed computing power scheduling layer matches one or more computing power usage requests for a computing power provider, the computing power provider can view the request details through the user interface, including the task type, estimated computing power consumption, expected revenue and computing power user credit score, and manually choose to accept or reject; only after the computing power provider manually confirms acceptance will the secure execution environment creation step be entered.

12. The method according to claim 1, characterized in that, The computing power provider can pre-set automatic acceptance conditions, including the minimum credit score threshold for acceptance, the maximum task unit price threshold for acceptance, or a whitelist of computing power users to be accepted; when a matching computing power usage request meets the automatic acceptance conditions, the system automatically accepts it without manual confirmation.

13. The method according to claim 1, characterized in that, The idle time period in the idle time determination condition is manually set and modified by the computing power provider through the user interface; the computing power provider can also set a daily maximum sharing time limit or a daily maximum revenue limit, and automatically stop sharing computing power externally after the limit is exceeded.

14. The method according to claim 1, characterized in that, When the edge device of the computing power user detects that the estimated computing power requirement of the current task exceeds the local available computing power threshold, or the local inference latency exceeds the preset limit, it will automatically trigger a request for computing power from outside, without the need for manual intervention by the user.

15. The method according to claim 1, characterized in that, Users of computing power can manually initiate external computing power requests through the user interface, manually or by voice inputting computing power requirements, including task type, estimated computing power consumption, and maximum acceptable price; the system automatically matches and recommends the best computing power nodes based on the user's input requirements, and the user can manually choose to accept or reject the recommendation results.

16. The method according to claim 1, characterized in that, Users of computing power can pre-set conditions for automatic external requests, including automatic requests when local computing power is insufficient, or automatic requests for specific types of tasks; other task types use manual confirmation mode by default; automatic and manual request modes can coexist, and users can switch flexibly according to task type.

17. The method according to claim 1, characterized in that, It also includes abnormal interruption and compensation steps: when a computing task is interrupted due to an abnormality in the computing power provider's equipment, the distributed computing power scheduling layer automatically migrates the unfinished task to other trusted computing power nodes for continued execution; additional costs incurred due to the interruption are determined and allocated by the distributed computing power scheduling layer based on the cause of the interruption.

18. The method according to claim 1, characterized in that, The computing power provider's edge computing device includes any one of personal AI host, smartphone, tablet, personal computer, and edge computing device; the hardware root of trust is at least one of an independent security chip, a hardware root of trust corresponding to a processor's built-in trusted execution environment, or a hardware security module that meets security certification standards.

19. The method according to claim 1, characterized in that, After the transaction settlement record is jointly signed by both parties' hardware trust roots, a storage file conforming to the legal electronic transaction certificate specifications is generated for tax declaration and financial audit retention.

20. The method according to claim 1, characterized in that, The digital signature and encrypted transmission are implemented using post-quantum cryptography algorithms, including at least one of the CRYSTALS-Kyber key encapsulation mechanism and the CRYSTALS-Dilithium digital signature algorithm.

21. The method according to claim 1, characterized in that, The distributed computing power scheduling layer automatically completes the donation transfer during transaction settlement according to the donation rules preset by the computing power provider, and generates an immutable donation record signed by the hardware root of trust, which serves as evidence for tax declaration and social responsibility report.

22. A hardware-based root of trust-based edge-side idle computing power sharing and secure transaction system, characterized in that, The device includes multiple edge computing devices with hardware roots of trust, a distributed computing power scheduling layer, and a computing power sharing agent module deployed inside each edge computing device, and is configured to perform the method described in any one of claims 1 to 21.

23. An edge computing device, characterized in that, It integrates a hardware root of trust and the computing power sharing agent module as described in claim 22.

Images