Algorithm-based credit scoring system and method based on decentralized continuous performance verification

Abstract

This invention provides an algorithmic credit scoring system and method based on decentralized continuous performance authentication. Through multi-party review, on-chain notarization, and proof-of-stake consensus, it constructs an algorithmic credit scoring ecosystem, achieving transparency, traceability, and high credibility in algorithm performance verification. The system includes: a notarization layer: a distributed ledger storing algorithm description information, cryptographic digests of algorithm execution resources and verification resources, and dynamically updated credit scores and change records; a resource layer: storing and providing executable algorithm implementation resources, verification programs containing standardized result output interfaces, verification datasets, and their descriptive information; a management layer participating in review management and task scheduling; and a verification layer employing a distributed verification network containing multiple computing nodes, executing: obtaining resource digests from the notarization layer and verifying resource integrity; executing verification programs in an independent environment to generate performance indicators; confirming valid verification results through a consensus mechanism; and a dynamic scoring engine adjusting credit scores in response to consensus results.

Description

Technical Field

[0001] This invention relates to the field of blockchain technology, and more specifically, to an algorithmic credit scoring system and method based on decentralized continuous performance authentication. Background Technology

[0002] In the fields of artificial intelligence, data science, and computational science, algorithm development is the core driving force behind technological progress. Whether it's theoretical innovation in academic research or engineering applications in industry, the effectiveness, robustness, and generalization ability of algorithms directly impact the value of technological achievements. Traditionally, algorithm performance verification relies on the following two methods:

[0003] The R&D team verifies the results themselves: providing limited test results in technical documents or papers;

[0004] Small-scale benchmarking: run evaluations on public datasets such as MNIST, subsets of ImageNet, etc.

[0005] Current mainstream verification methods have systemic flaws, mainly reflected in:

[0006] Insufficient credibility: The verification process is led by the algorithm proposer and lacks independent third-party supervision, which poses the risk of selective data disclosure or embellishment of results, such as only showing the best case.

[0007] Limited coverage: The test dataset is small in scale and the scenario is simple, making it difficult to reflect the algorithm's real performance in complex environments, such as data distribution deviation and heterogeneous hardware.

[0008] Untraceability: The verification process, raw data, and operating environment are not fully recorded, making it impossible to independently reproduce or audit the results;

[0009] Limitations of static evaluation: One-time verification cannot capture the data adaptability of an algorithm as it evolves, such as the concept drift problem.

[0010] To enhance credibility, some fields have attempted to introduce the following mechanisms: third-party certification: test reports are issued by independent laboratories, but this suffers from opaque processes and inconsistent standards; open-source reproduction communities: peer verification is encouraged through open code sharing, but participation relies on voluntary participation and lacks incentives and constraints. This leads to inconsistent quality of academic achievements and low efficiency in industrial transformation, forming a key bottleneck in the technological innovation chain. Summary of the Invention

[0011] The purpose of this invention is to provide an algorithmic credit scoring system and method based on decentralized continuous performance authentication, which can simultaneously satisfy: transparency, with the entire verification process being auditable; dynamism, supporting continuous environmental adaptation assessment; trustworthiness, with a tamper-proof multi-party consensus mechanism; and scalability, with a unified evaluation framework for cross-domain algorithms.

[0012] The embodiments of the present invention are implemented as follows:

[0013] An algorithmic credit scoring system based on decentralized continuous performance authentication, comprising:

[0014] The blockchain-based evidence storage module is used for permanent storage.

[0015] Algorithm metadata includes algorithm name, version, release date, scope of application, and document hash;

[0016] Verifiable resource packages include approved algorithm code, model files, validation scripts, dataset specifications and their corresponding hash values;

[0017] Dynamic credit scoring records, including initial scores, ongoing verification results, review comments, and score change logs;

[0018] The decentralized verification network module, composed of multiple geographically distributed nodes, executes:

[0019] Obtain the hash value of the verifiable resource package from the blockchain and verify the consistency between the locally downloaded resource and the on-chain hash;

[0020] Run the verification script in one or more heterogeneous environments to generate a result summary containing performance metrics and its hash value;

[0021] A valid consensus is achieved by cross-validating the result summaries from multiple nodes through a proof-of-stake consensus mechanism.

[0022] The consensus-confirmed verification result triggers the smart contract to update the algorithm's credit score;

[0023] The dynamic scoring engine adjusts credit scores in real time based on the following rules:

[0024] When the deviation of the continuous verification performance index from the initial value is less than or equal to the dynamic threshold, the score is increased to the first preset value.

[0025] When the deviation is greater than the dynamic threshold, the score is reduced by the second preset value, and |second preset value| > |first preset value|;

[0026] The dynamic threshold is adaptively calculated based on the algorithm type and performance characteristics.

[0027] In a preferred embodiment of the present invention, the above-mentioned proof-of-stake consensus mechanism specifically includes:

[0028] Verification nodes are assigned voting weights based on the amount of staked tokens and their historical verification accuracy.

[0029] Consensus must be reached to meet the following conditions:

[0030] The hashes of the result digests submitted by more than 70% of the weighted nodes are consistent.

[0031] The number of participating nodes is greater than or equal to the minimum number of safe nodes.

[0032] If no consensus is reached, the smart contract will automatically schedule a high-reputation node to initiate the dispute re-verification process.

[0033] In a preferred embodiment of the present invention, the generation of the verifiable resource package described above must satisfy the following:

[0034] The algorithm code and verification script are stored separately, and the script contains standardized input / output interfaces;

[0035] The dataset specifications clearly define the data source, collection time, and preprocessing methods;

[0036] All files within the resource package will generate a Merkle tree root hash and store it on the blockchain.

[0037] In a preferred embodiment of the present invention, the dynamic scoring engine further performs:

[0038] Based on the continuous verification consistency rate, the developer credit score of the algorithm proposer is adjusted synchronously.

[0039] The reviewer's review credit score will be adjusted based on the degree of consistency between the review comments and the results of ongoing verification.

[0040] The scoring adjustment record is linked to the hash value of the corresponding verification result and uploaded to the blockchain.

[0041] In a preferred embodiment of the present invention, the verification triggering conditions of the decentralized verification network module include:

[0042] Scheduled trigger: Automatically retrieve new datasets and perform verification at preset intervals;

[0043] Event-driven: Triggered when a change in key parameters of the algorithm's applicable scenario is detected.

[0044] This invention also provides an algorithmic credit scoring method based on decentralized continuous performance authentication, which, based on any of the aforementioned algorithmic credit scoring systems, includes the following steps:

[0045] Trusted initialization: The algorithm proposer submits a verifiable resource package and algorithm metadata package containing preliminary verification results; a third-party reviewer independently reproduces the verification process and generates a review report that includes executability, scientific validity, and consistency of reproduction; the system assigns an initial score based on the review report; the smart contract packages the resource package, review report, and initial score to generate the genesis block.

[0046] Continuous effectiveness verification involves a decentralized verification network performing multiple rounds of verification within a time window [T1, T2]. Verification methods include:

[0047] The node retrieves the latest resource package hash from the chain, downloads the resource, and verifies its integrity.

[0048] Run the verification script in the local environment and output performance metrics with timestamps.

[0049] Each node submits a summary of performance metrics to the consensus layer;

[0050] The effective indicator value is confirmed through the proof-of-stake consensus.

[0051] The dynamic scoring engine compares the deviation Δ between the effective indicator value and the initial value, and updates it according to the following rules:

[0052] If the deviation Δ ≤ dynamic threshold δ, then the score new =Score old+ k1;

[0053] If the deviation Δ > the dynamic threshold δ, then the score new =Score old -k2, where k2>k1>0;

[0054] End-to-end traceability: All verification records, scoring operations, and participant behaviors generate a publicly auditable traceability chain; the scoring trend and the distribution of multi-node verification results are dynamically displayed through a visual interface.

[0055] In a preferred embodiment of the present invention, the dynamic threshold δ is calculated as follows:

[0056] Where: β is the algorithm robustness coefficient (0 < β ≤ 1); σ is the standard deviation of historical validation results; μ is the mean of historical validation results; I o This is the initial verification value.

[0057] In a preferred embodiment of the present invention, the above consensus confirmation also includes:

[0058] If performance metrics exhibit abnormal fluctuations during a single round of verification:

[0059] Automatically trigger cross-regional node verification;

[0060] The hash of the retest result is compared with that of the original result;

[0061] If there is a discrepancy, the algorithm score will be frozen and a manual audit process will be initiated.

[0062] In a preferred embodiment of the present invention, the above-mentioned end-to-end tracing includes:

[0063] To query any verification event using a blockchain explorer, perform the following steps:

[0064] Associated resource package Merkle tree path;

[0065] Participating node reputation weight distribution;

[0066] Score adjustment decision tree log;

[0067] Supports multi-dimensional filtering and verification of records by time, geographic location, and dataset version.

[0068] An electronic device includes a processor and a memory, the memory storing executable instructions; the processor executes the instructions to implement the steps of any of the methods described above, and:

[0069] The device operates as a consensus node in a decentralized verification network;

[0070] Encrypted fragmentation of local storage resource layer data;

[0071] The verification script is executed in isolation from the host operating system.

[0072] The beneficial effects of the embodiments of the present invention are:

[0073] 1. Provide a trusted authentication system, which adopts blockchain full-process document / code / result hash storage, timestamp traceability chain, and multi-party consensus anti-tampering collaboration to solve the problem of lack of credibility in algorithm verification, and establish a quantifiable trust mechanism for academic / industrial applications;

[0074] 2. The system adopts a dynamic evaluation mechanism, which uses multiple rounds of timed / event / manual triggering verification and asymmetric scoring rules, such as small rewards and large penalties, and adaptive deviation thresholds, to truly reflect the robustness of the algorithm in evolving data and avoid distortion in static evaluation.

[0075] 3. The system adopts a standardized verification framework, unifying metadata format, standardizing verification script interface, and mandating environment declaration and output specifications to achieve cross-platform and cross-node result comparability and eliminate interference from human and environmental differences;

[0076] 4. The system features decentralized collaboration, employing a node admission mechanism, proof-of-stake consensus, and anti-collusion verification design to enhance the system's resilience against single points of failure, prevent malicious manipulation, and ensure network fairness.

[0077] 5. The system's ecosystem incentive loop uses tokens to incentivize high-quality verification nodes, dynamic reputation scores for developers / reviewers / nodes, and resource allocation linked to credit scores to drive continuous optimization and form a positive cycle of "improved verification quality - algorithm improvement - ecosystem value growth". Attached Figure Description

[0078] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0079] Figure 1 This is a schematic block diagram of the algorithm credit scoring system according to Embodiment 1 of the present invention;

[0080] Figure 2 This is the decentralized verification three-stage process of Embodiment 1 of the present invention;

[0081] Figure 3 This is a flowchart of the algorithm credit scoring method according to Embodiment 2 of the present invention;

[0082] Figure 4 This is a flowchart of the credit scoring algorithm for nighttime visible light satellite image inversion in Embodiment 3 of the present invention;

[0083] Figure 5 This is a flowchart of the full-link credit scoring verification process for the nighttime visible light satellite image inversion algorithm in Embodiment 4 of the present invention;

[0084] Figure 6 This is a flowchart of the credit scoring algorithm for the nighttime visible light satellite image inversion algorithm of Embodiment 5 of the present invention;

[0085] Figure 7 This is a schematic block diagram of an electronic device that executes the algorithm credit scoring method in Embodiment Six of the present invention.

[0086] Icons: Algorithm credit scoring system 100; Blockchain evidence storage module 110; Decentralized verification network module 120; Data retrieval 121; Local verification 122; Consensus confirmation 123; Dynamic scoring engine 130; Evidence storage layer 101; Resource layer 102; Management layer 103; Verification layer 104; Electronic device 200; Processor 210; Memory 220. Detailed Implementation

[0087] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0088] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0089] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0090] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art.

[0091] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0092] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0093] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0094] First embodiment

[0095] Please see Figure 1-2 This embodiment provides an algorithmic credit scoring system 100 based on decentralized continuous effectiveness authentication. Its practical application areas include, but are not limited to: academic research and paper publication, technology achievement display and promotion, technology achievement transaction and evaluation, technology achievement transformation and industrialization, and technology achievement engineering application and implementation.

[0096] It includes a blockchain evidence storage module 110, a decentralized verification network module 120, and a dynamic scoring engine 130, constructing a three-in-one trust guarantee of process trust (blockchain evidence storage), result trust (multi-node consensus), and subject trust (dynamic reputation scoring).

[0097] The blockchain evidence storage module 110 stores algorithm metadata and verifiable resource packages involved in the evidence storage layer 101, resource layer 102, management layer 103, and verification layer 104, and permanently stores dynamic credit scoring records. The entire process is tamper-proof, enhancing credibility. Using blockchain technology, all verification steps related to the algorithm, including algorithm documentation, source code, verification procedures, dataset descriptions, review comments, and each verification result, are recorded on the chain in the form of hash values. All historical records are unmodifiable or undeletable, preventing data tampering, forgery, or loss, significantly improving the authority and credibility of the algorithm verification and scoring process.

[0098] Among them, the evidence storage layer 101 involves blockchain and smart contracts.

[0099] Blockchain network: As the core infrastructure of distributed ledger technology, it provides immutable, traceable, and permanent storage for recording verification results and credit assessment scores at each stage of the algorithm's lifecycle.

[0100] Smart contracts: Execute the evidence storage logic in an automated and programmable manner, and are responsible for managing the on-chain publishing, querying, and fine-grained access control of the following key operations:

[0101] Algorithm metadata: Algorithm description documents and their metadata, including algorithm name, version number, release date, and applicable scenarios.

[0102] Verifiable resource package: algorithm implementation code, verification script, dataset description document and its corresponding cryptographic hash value.

[0103] Dynamic credit scoring records include: preliminary algorithm validation results, expert review opinions, periodic continuous validation records, and changes in credit scores.

[0104] Resource layer 102 serves as an algorithm and verification resource library, and can be categorized as verifiable resource package data for:

[0105] Standardized asset storage, which includes: audited and verified executable code, pre-trained model files and their cryptographic hashes;

[0106] Verification resource archiving: Systematically manage verification process scripts, dataset specification documents and their corresponding cryptographic hash values ​​to ensure that the verification process is reproducible;

[0107] Flexible deployment architecture: Supports centralized and distributed storage architectures, adapting to diverse network infrastructure deployment needs.

[0108] The management layer 103 is used for reviewers and task scheduling. It has an internal reviewer information database to record the basic information and expertise of reviewers and dynamically allocate review tasks. The management layer 103 is equipped with a task scheduler that triggers preliminary and continuous verification and credit scoring update tasks based on time or events, and coordinates execution nodes.

[0109] The decentralized verification network module 120, composed of multiple geographically distributed nodes, is used to construct the decentralized verification network of the verification layer 104. In the decentralized verification network, any authorized node can participate in algorithm verification and credit scoring contributions, providing computing power and storage resources.

[0110] For details, please see Figure 2 The decentralized verification network module 120 performs the following:

[0111] Data retrieval 121: Obtain verifiable resource package hash values ​​from the smart contract of the blockchain network, including at least the package hash, verification script hash, and dataset hash, and verify the consistency between the locally downloaded resources and the on-chain hashes. The locally downloaded resources include the corresponding packages, scripts, and datasets downloaded from resource layer 102.

[0112] Local Validation 122: Run the validation script in one or more heterogeneous environments, run the algorithm based on the downloaded dataset, and generate a result summary containing performance metrics and its hash value;

[0113] Consensus Confirmation 123: Cross-validate the result summaries of multiple nodes through the Proof-of-Stake consensus mechanism to reach a valid consensus; trigger the smart contract to update the algorithm credit score by confirming the consensus verification result. A predetermined consensus algorithm can be used, such as comparing the result summaries based on majority or Proof-of-Stake. Once consistent, a verification consensus is reached and the algorithm credit score is updated.

[0114] The dynamic scoring engine 130 writes the consensus-confirmed verification results and their hashes to the blockchain. Based on node contributions and verification accuracy, the smart contract automatically distributes incentives and adjusts the algorithm's credit score. Specifically, the credit score is adjusted in real time according to the following rules:

[0115] When the deviation of the continuous verification performance index from the initial value is less than or equal to the dynamic threshold δ, the score is increased by the first preset value k1.

[0116] When the deviation is greater than the dynamic threshold δ, the score is reduced by the second preset value k2, and |second preset value|>|first preset value|. Using an asymmetric scoring function, if the deviation is small, the system provides a small reward; if the deviation is large, the system imposes a large penalty, which truly reflects the robustness of the algorithm in evolving data. In this embodiment, the dynamic threshold δ is adaptively calculated based on the algorithm type and performance index characteristics.

[0117] In this embodiment, the proof-of-stake consensus mechanism specifically includes: verification nodes are assigned voting weights based on the amount of staked tokens and historical verification accuracy; consensus is reached when: the hashes of the result digests submitted by more than 70% of the weighted nodes are consistent; the number of participating nodes is greater than or equal to the minimum number of secure nodes; if consensus is not reached, the smart contract automatically schedules high-reputation nodes to initiate the dispute re-verification process.

[0118] The system is designed with incentive mechanisms such as tokens or points to reward nodes that continuously participate and contribute high-quality verification results, thereby increasing community activity and node participation. Simultaneously, the credit score of the algorithm proposer will dynamically change based on verification performance, incentivizing algorithm authors to continuously optimize their work and promoting healthy industry development and fair competition.

[0119] In this embodiment, the generation of the verifiable resource package must meet the following requirements: the algorithm code and the verification script are stored separately, and the script contains standardized input and output interfaces; the dataset specifications clearly define the data source, collection time, and preprocessing method; and all files in the resource package generate Merkle tree root hashes for on-chain storage.

[0120] It boasts advantages in standardized verification and strong result comparability, clearly defining algorithm metadata formats, verification script interfaces, and dataset description standards. It unifies the specifications for algorithm submission, review, and verification materials, ensuring a high degree of consistency in execution conditions across different verification nodes. Results from different nodes can be evaluated using objective and comparable metrics such as SSIM, accuracy, and recall, eliminating fluctuations caused by environmental and human factors. Furthermore, the standardized scripts can be quickly migrated to new application scenarios when verification scripts are changed or datasets are updated, achieving interoperability between heterogeneous systems.

[0121] In this embodiment, the dynamic scoring engine 130 also performs the following: adjusting the developer credit score of the algorithm proposer synchronously based on the continuous verification consistency rate; adjusting the reviewer's review credit score based on the consistency between the review opinions and the continuous verification results; and binding the score adjustment record to the hash value of the corresponding verification result on the blockchain.

[0122] Employing dynamic and continuous evaluation, the system accurately reflects algorithm performance and supports multiple rounds of continuous verification, whether periodic, event-driven, or user-initiated. Each round's verification results dynamically impact the algorithm's credit score. As new data and application scenarios emerge, the system dynamically adjusts the algorithm's credit score, ensuring the score remains synchronized with the algorithm's true capabilities and enhancing the practical significance of algorithm performance evaluation.

[0123] In this embodiment, the verification triggering condition of the decentralized verification network module 120 can be the following:

[0124] The system is triggered on a timed basis, automatically pulling new datasets for verification according to a preset cycle. Verification is performed periodically at fixed time intervals through the management layer's 103 timed task scheduling.

[0125] Event-driven, triggered when changes in key parameters of the algorithm's applicable scenario are detected, can be achieved by listening to on-chain events through smart contracts;

[0126] Manual requests can be made by triggering verification through staking tokens when there are situations such as disputed verification or targeted assessment.

[0127] Second embodiment

[0128] Please see Figure 3 This invention provides an algorithmic credit scoring method based on decentralized continuous performance authentication, and based on the algorithmic credit scoring system described in the first embodiment, includes the following steps:

[0129] S1: Trusted initialization. The algorithm proposer submits a verifiable resource package and algorithm meta-data package containing preliminary verification results. After conducting preliminary verification on a small dataset, the algorithm proposer submits the algorithm documentation, source code, verification script, dataset, and preliminary results.

[0130] S2: Preliminary review, third-party reviewers independently reproduce and verify the process, and generate a review report that includes feasibility, scientific validity and consistency of reproduction;

[0131] S3: On-chain evidence storage and initial scoring. The system packages and uploads the algorithm, verification materials, and review opinions to the blockchain through smart contracts. It allocates initial scores and records the initial credit scores based on the review reports. The smart contract packages the resource package, review reports, and initial scores to generate the genesis block.

[0132] S4: Continuous verification and scoring updates, multi-party verifier continuous effectiveness certification, decentralized verification network performs multiple rounds of repeated verification within the time window [T1, T2], verification methods include:

[0133] The node retrieves the latest resource package hash from the chain, downloads the resource, and verifies its integrity.

[0134] Run the verification script in the local environment and output performance metrics with timestamps.

[0135] Each node submits a summary of performance metrics to the consensus layer;

[0136] The effective indicator value is confirmed through the proof-of-stake consensus.

[0137] The dynamic scoring engine compares the deviation Δ between the effective indicator value and the initial value, and updates it according to the following rules:

[0138] If the deviation Δ ≤ dynamic threshold δ, then the score new =Score old+ k1;

[0139] If the deviation Δ > the dynamic threshold δ, then the score new =Score old -k2, where k2>k1>0.

[0140] The decentralized network periodically or in an event-driven manner repeats the verification process, stores the results on the blockchain, and dynamically updates the algorithm credit score based on verification consistency and frequency, thereby achieving continuous effectiveness certification that is traceable and comparable across multiple time points and environments.

[0141] This algorithmic credit scoring method also enables end-to-end traceability. All verification records, scoring operations, and participant behaviors generate a publicly auditable traceability chain, facilitating auditing and reproduction. This invention automatically generates on-chain timestamps and detailed operation records for each algorithm verification, scoring adjustment, and node participation, forming a complete verification traceability chain. Whether it's regulatory authorities, academic institutions, or industry users, they can track the algorithm's historical verification path at any time, facilitating problem tracing, dispute resolution, and independent reproduction of the verification process, greatly improving the system's compliance and auditability. Finally, a visual interface dynamically displays scoring trends and the distribution of multi-node verification results.

[0142] In this embodiment, the dynamic threshold δ is calculated as follows: Where: β is the algorithm robustness coefficient (0 < β ≤ 1); σ is the standard deviation of historical validation results; μ is the mean of historical validation results; I o This is the initial verification value.

[0143] In this embodiment, consensus confirmation also includes the following operations: if the performance index fluctuates abnormally in a single round of verification, cross-regional node re-verification is automatically triggered; the re-verification result is compared with the original result hash, and if they are inconsistent, the algorithm score is frozen and a manual audit process is initiated.

[0144] In this embodiment, end-to-end traceability also includes: querying any verification event through a blockchain explorer and executing the following steps: associating the resource package Merkle tree path; distributing the reputation weights of participating nodes; adjusting the score adjustment decision tree log; and supporting multi-dimensional filtering of verification records by time, geographical location, and dataset version. This enhances system stability and resilience. Any eligible node can participate in algorithm verification, result consensus, and on-chain notarization. The system has no single point of failure and can effectively defend against risks such as hacker attacks and service interruptions. Multi-node collaborative verification also prevents malicious tampering or fraud by individual nodes, ensuring the fairness and stability of the entire scoring and verification network.

[0145] In summary, the algorithm credit scoring method provided in this embodiment helps improve the quality of academic achievements and facilitates industrial technology transformation. The algorithm credit scoring system provides the academic community with a quantifiable and traceable tool for evaluating algorithm quality, improving the efficiency and credibility of academic paper review. Simultaneously, industry can use this method to quickly screen high-scoring, robust, and high-quality algorithms, significantly reducing the threshold and cost of evaluating and introducing new technologies, and accelerating the transformation of scientific research results into practical productivity. The system is flexible and scalable, compatible with diverse algorithms and application scenarios. This system supports algorithm verification of different types and fields, applicable to traditional machine learning, deep learning, signal processing, remote sensing image analysis, and more. Through flexible configuration of verification scripts and datasets, the system can be expanded to more technical fields and cross-industry application scenarios, possessing high openness and compatibility.

[0146] Third Embodiment

[0147] To illustrate the specific application effects of the system of this invention, the complete credit scoring process of the "Nighttime Visible Light Satellite Image Inversion Algorithm" is used as an example. Please refer to [link / reference]. Figure 4 The detailed description is as follows:

[0148] 1. Algorithm Proposal and Preliminary Validation

[0149] Research team A developed an innovative algorithm for reconstructing nighttime visible light satellite imagery using infrared data. The algorithm aims to reconstruct nighttime images with a high structural similarity to daytime visible light satellite images using infrared observation data from the Himawari-8 satellite. Team A first selected infrared observation data from the night of May 1, 2024, as input, and used actual daytime visible light satellite images obtained on the same day as a reference. A quantitative comparison was performed using the Structural Similarity Index (SSIM), achieving a good performance of SSIM=0.868. To ensure the objectivity and reproducibility of the algorithm verification, Team A packaged the algorithm source code, verification scripts, original dataset, experimental procedures, and all preliminary verification results according to the system specifications and submitted them to the designated third-party reviewer, B.

[0150] 2. Third-party review and initial credit score

[0151] After receiving the task assignment from the system, reviewer B first verified the hash values ​​of the algorithm package and all files, unpacked the verification materials, and independently reproduced the algorithm's operation. B focused on reviewing the completeness and executability of the source code, checking whether the verification script was consistent with the documented process, and confirming that the selection and processing logic of the experimental data was scientifically sound. Simultaneously, B independently ran the preliminary verification process, comparing the SSIM calculation results with those reported by verification team A. The data matched the actual output, with no false or misleading information. After confirming compliance with scientific methods and standards, B issued a detailed review report, confirming that the algorithm possesses high technical innovation and engineering feasibility. Based on the review conclusions, the system automatically uploaded all review materials and verification results to the blockchain via a smart contract, assigning an initial credit score (e.g., 0.75 / 1.0) to the algorithm.

[0152] 3. Account Distribution and Credit Score Inquiry

[0153] The system assigns a dedicated account to Team A, allowing the team to log in at any time to view the algorithm's on-chain evidence records, detailed review comments, historical and latest credit scores, and a complete record of all subsequent ongoing verifications. This feature helps the algorithm proposer understand changes in the algorithm's credibility within the industry and make targeted adjustments for future optimization.

[0154] 4. Decentralized continuous verification and dynamic scoring mechanism

[0155] As the algorithm enters the open verification phase, several authorized third-party continuous verifiers (C) can voluntarily or as scheduled by the system to participate in subsequent verifications. Each month, C downloads the latest "Sunflower 8" satellite infrared data from official channels and automatically obtains the on-chain proven algorithm package and verification scripts. On the local verification node, C verifies the software package, scripts, and dataset hash values ​​one by one to ensure that the data used is completely consistent with the on-chain proven version. Subsequently, C runs the algorithm and uses the visible light satellite imagery of the day as a reference to recalculate SSIM. The verification output and its hash value are then uploaded to the blockchain via a smart contract.

[0156] The system automatically compares the newly generated SSIM value with the initially recorded SSIM=0.868. If the difference between the newly verified SSIM and the original result is 0.02 (i.e., 0.848), which does not exceed the dynamic threshold set by the system, the system determines that this round of verification has passed, the algorithm's credit score is increased by 0.02 points, and the relevant nodes and verification summary are recorded. If the difference is greater than 0.02, which exceeds the dynamic threshold set by the system, the algorithm is determined to have some instability or insufficient robustness, the credit score is decreased by 0.05 points accordingly, and the abnormal verification event is marked on the chain.

[0157] 5. Transparent and traceable results for multiple parties

[0158] The algorithm proposer (A), all continuous validators (C), industry users, academic institutions, and regulatory authorities can all use the system interface to search and browse the results of all continuous validations of the algorithm at any time. Queries include validation summaries for each node, raw data hashes, SSIM metrics, dynamic changes in scores, and the traceability path of all on-chain evidence files. The system supports screening by multiple dimensions such as time, node, and dataset, facilitating rapid location and analysis of the algorithm's performance in different scenarios, and comprehensively ensuring the openness, transparency, and traceability of the algorithm's performance.

[0159] Based on this process, industry users can quickly determine the practicality and reliability of an algorithm based on its credit score, continuous stability verification, and historical verification data, greatly reducing the decision-making risks caused by "black box" algorithms or data fabrication. The academic community can use this to evaluate the contribution of papers, promoting the publication and reuse of high-quality results; while enterprises can use scores to select and adopt the optimal algorithm, accelerating the commercialization of research findings.

[0160] Fourth embodiment

[0161] To facilitate the explanation of the practical application process of this invention, the full-link credit scoring verification process of the "Nighttime Visible Light Satellite Image Inversion Algorithm" is used as an example. The process is similar to that in the third embodiment, but there are some differences. Please refer to [link to relevant documentation]. Figure 5 The details are as follows:

[0162] 1. Algorithm Proposal and Preliminary Validation

[0163] Researcher A proposed an algorithm for retrieving nighttime visible light satellite imagery based on infrared light data. A first completed the algorithm design, implementation, and preliminary experiments in an independent research environment, using infrared remote sensing data from the "Sunflower 8" satellite on the night of May 1, 2024, to reconstruct nighttime visible light imagery through a specific algorithmic process. To quantify the algorithm's performance, A selected real daytime visible light satellite imagery from the same day as a benchmark and used the Structure Similarity Index (SSIM) for objective evaluation, achieving an excellent score of 0.868. To ensure reproducibility, A compiled and packaged the algorithm source code, detailed validation scripts, original dataset, data processing flow, experimental parameters, and all validation outputs, strictly adhering to the system format to generate hash values, and submitted them to designated reviewer B, forming a complete preliminary validation material package.

[0164] 2. Review and Initial Scoring

[0165] The system automatically assigns the task to third-party reviewer B. Upon receiving the material package, B verifies the hash value of each file to ensure the data has not been tampered with. Based on the submitted documents and script instructions, B independently reproduces the algorithm's execution in a local environment. Comparing it with A's experimental process, B confirms that the data processing at each step conforms to scientific standards, the code has no logical flaws, the verification process is complete, and the resulting SSIM value is consistent with A's report. Reviewer B writes a detailed review report, including evaluations of the algorithm's innovativeness, scientific validity, and engineering reproducibility, and uploads the review comments through the system. The smart contract automatically stores the algorithm package, all verification data, review report, reviewer and algorithm author identity hashes on the blockchain for unified notarization, and assigns A an initial algorithm credit score of 0.75 (out of 1.0) based on the review conclusions.

[0166] 3. Account Distribution and Inquiry

[0167] Unlike the third embodiment, the system generates blockchain identities and accounts for both A and B, and grants them permissions. This allows A to log in to the platform at any time to query all evidence storage information, initial credit scores, review comments, and continuous verification records of the algorithm in real time. As a reviewer, B can also query the evidence storage status of the projects they participate in, the algorithm's scoring history, and the comparison results of each verification, facilitating their continued participation in continuous verification or being invited for further review.

[0168] 4. Continuous verification and rating updates

[0169] As the algorithm goes live, the system automatically opens a continuous verification channel. The intelligent scheduler assigns several third-party continuous verifiers, such as C1, C2, and C3. Verifiers can participate in decentralized verification from different geographical locations or computing nodes. Verifiers automatically download the on-chain stored algorithm package, verification script, and parameter descriptions through the system's intelligent scheduling, and verify that their hash values ​​are completely consistent with the on-chain registered values. Every 10 minutes, the system periodically pulls the latest infrared observation data from the "Sunflower 8" satellite from official channels and uses it, along with on-chain visible light comparison data, to run the verification process locally. The verification process includes:

[0170] (a) Automatically compare the downloaded algorithm package, script, and dataset hash with the on-chain evidence to prevent tampering and forgery;

[0171] (b) Run the algorithm to reconstruct nighttime visible light images, calculate the SSIM index in real time, and automatically save all intermediate outputs, logs, and result files;

[0172] (c) Package the new SSIM value, the identity of the verification node, the hash results of the verification data and the running log through a smart contract, store them on the blockchain, and form a complete multi-node, multi-time point verification chain.

[0173] The system compares the deviation between the latest SSIM and the initial 0.868. If the deviation is no greater than 0.02 (i.e., SSIM 0.848), the algorithm is deemed to have consistently performed stably, and the current credit score is increased by 0.02 points; if the deviation exceeds this, the algorithm is considered unstable, and the credit score is decreased by 0.05 points. Simultaneously, the R&D and review credit scores of A and B are dynamically adjusted based on validation consistency to ensure that the overall reputation of the algorithm and participating parties is fairly reflected in the system.

[0174] The system provides a visual interface for stakeholders such as A, B, all validators, industry users, and academia. It dynamically displays all continuous verification results of the algorithm at different nodes and times, historical SSIM metric curves, the identity and runtime summary of each verification node, hash verification results of all raw data and the runtime environment, and dynamics of initial review and continuous scoring. The system also supports exporting all verification and scoring history, anomaly detection and alarms, and targeted invitations to new nodes to participate in reproduction verification. This truly achieves full-process, full-node, and all-time traceability and transparent monitoring of algorithm performance, providing strong data support and decision-making backing for industry, academia, regulatory authorities, and other stakeholders.

[0175] Fifth embodiment

[0176] Please see Figure 6 To illustrate the application effect of this invention, the nighttime visible light satellite image inversion algorithm is used as an example:

[0177] 1. Algorithm Proposal and Preliminary Validation

[0178] Researcher A published an algorithm for retrieving nighttime visible light satellite imagery based on infrared data, claiming that the algorithm can generate nighttime visible light images comparable to daytime visible light images using infrared data from the Himawari-8 satellite. Using infrared data acquired by Himawari-8 on the night of May 1, 2024, A achieved a Structural Similarity Index (SSIM) score of 0.868 on the corresponding daytime visible light image. A submitted the algorithm's source code, validation scripts, the dataset used, and preliminary validation results to reviewer B.

[0179] 2. Review and Initial Scoring

[0180] Reviewer B reviewed the submitted materials, focusing on verifying the executability of the source code, the completeness of the verification process, and the rationality of the data processing. B confirmed that A's experimental method conformed to scientific methodology and that the verification results were credible. After the review was approved, the system used a smart contract to store the algorithm and verification results on the blockchain, and assigned A an initial algorithm credit score of 0.75 (out of 1.0).

[0181] 3. Account Distribution and Inquiry

[0182] The system distributes personal accounts to A and B, who can log in to the system to view their algorithm credit scores and verification records for each round in real time.

[0183] 4. Continuous verification and rating updates

[0184] Several third-party continuous verifiers, C1, C2, and C3, download the algorithm software package, execution script, and verification script. Every 10 minutes, they obtain the latest Sunflower 8 infrared light data from official channels and execute the verification process locally.

[0185] (a) Verify that the hash values ​​of the software package and the running script are consistent with those on the chain;

[0186] (b) Run the algorithm and calculate the new SSIM value;

[0187] (c) Upload the verification results and their hash to the blockchain via a smart contract.

[0188] The system automatically compares the newly generated SSIM with the initial 0.868: if the deviation is 0.02, the verification is considered successful, and the algorithm's credit score is increased by 0.02; otherwise, the verification is considered unsuccessful, and the credit score is decreased by 0.05. The R&D credit scores of A and B will be dynamically adjusted based on the verification results.

[0189] 5. Results Display

[0190] A and other stakeholders can view the dynamic changes in the SSIM value of each continuous verification, the verification summary of each node, and the algorithm credit score on the system interface, enabling full traceability and transparent monitoring of algorithm performance.

[0191] Sixth Embodiment

[0192] Please see Figure 7 This embodiment provides an electronic device 200, including one or more processors 210 and a memory 220, the memory storing executable instructions; when the processor executes the instructions, it implements the steps of an algorithmic credit scoring method, and:

[0193] The device operates as a consensus node in a decentralized verification network;

[0194] Encrypted fragmentation of local storage resource layer data;

[0195] The verification script is executed in isolation from the host operating system.

[0196] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0197] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, or may be electrical, mechanical, or other forms of connection.

[0198] The units described as separate components may or may not be physically separate. As will be appreciated by those skilled in the art, the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the present invention.

[0199] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0200] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or grid device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0201] The above specific embodiments further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An algorithmic credit scoring system based on decentralized continuous performance authentication, characterized in that, include: The blockchain-based evidence storage module is used for permanent storage. Algorithm metadata includes algorithm name, version, release date, scope of application, and document hash; Verifiable resource packages include approved algorithm code, model files, validation scripts, dataset specifications and their corresponding hash values; Dynamic credit scoring records, including initial scores, ongoing verification results, review comments, and score change logs; The decentralized verification network module, composed of multiple geographically distributed nodes, executes: Obtain the hash value of the verifiable resource package from the blockchain and verify the consistency between the locally downloaded resource and the on-chain hash; Run the verification script in one or more heterogeneous environments to generate a result summary containing performance metrics and their hash values; A valid consensus is achieved by cross-validating the result summaries from multiple nodes through a proof-of-stake consensus mechanism. The consensus-confirmed verification result triggers the smart contract to update the algorithm's credit score; The dynamic scoring engine adjusts credit scores in real time based on the following rules: When the deviation of the continuous verification performance index from the initial value is less than or equal to the dynamic threshold, the score is increased to the first preset value. When the deviation is greater than the dynamic threshold, the score is reduced by the second preset value, and |second preset value| > |first preset value|; The dynamic threshold is adaptively calculated based on the algorithm type and performance characteristics. The aforementioned proof-of-stake consensus mechanism specifically includes: Verification nodes are assigned voting weights based on the amount of staked tokens and their historical verification accuracy. Consensus must be reached to meet the following conditions: The hashes of the result digests submitted by more than 70% of the weighted nodes are consistent. The number of participating nodes is greater than or equal to the minimum number of safe nodes. When no consensus is reached, the smart contract automatically schedules a high-reputation node to initiate the dispute re-verification process; The dynamic scoring engine also performs: Based on the continuous verification consistency rate, the developer credit score of the algorithm proposer is adjusted synchronously. The reviewer's review credit score will be adjusted based on the degree of consistency between the review comments and the results of ongoing verification. The scoring adjustment record is linked to the hash value of the corresponding verification result and uploaded to the blockchain.

2. The algorithmic credit scoring system based on decentralized continuous performance authentication according to claim 1, characterized in that, The generation of the verifiable resource package must meet the following requirements: The algorithm code and verification script are stored separately, and the script contains standardized input / output interfaces; The dataset specifications clearly define the data source, collection time, and preprocessing methods; All files within the resource package will generate a Merkle tree root hash and store it on the blockchain.

3. The algorithmic credit scoring system based on decentralized continuous performance authentication according to claim 1, characterized in that, The verification triggering conditions of the decentralized verification network module include: Scheduled trigger: Automatically retrieve new datasets and perform verification at preset intervals; Event-driven: Triggered when a change in key parameters of the algorithm's applicable scenario is detected.

4. An algorithmic credit scoring method based on decentralized continuous performance authentication, based on the algorithmic credit scoring system described in any one of claims 1-3, characterized in that, Includes the following steps: Trusted initialization: The algorithm proposer submits a verifiable resource package and algorithm metadata package containing preliminary verification results; a third-party reviewer independently reproduces the verification process and generates a review report that includes executability, scientific validity, and reproducibility consistency; the system assigns an initial score based on the review report; the smart contract packages the resource package, review report, and initial score to generate the genesis block. Continuous effectiveness verification involves a decentralized verification network performing multiple rounds of verification within a time window [T1, T2]. Verification methods include: The node retrieves the latest resource package hash from the chain, downloads the resource, and verifies its integrity. Run the verification script in the local environment and output performance metrics with timestamps. Each node submits a summary of performance metrics to the consensus layer; The effective indicator value is confirmed through the proof-of-stake consensus. The dynamic scoring engine compares the deviation Δ between the effective indicator value and the initial value, and updates it according to the following rules: If the deviation Δ ≤ dynamic threshold δ, then the score new =Score old+ k1; If the deviation Δ > the dynamic threshold δ, then the score new =Score old -k2, where k2>k1>0; End-to-end traceability: All verification records, scoring operations, and participant behaviors generate a publicly auditable traceability chain; the scoring trend and the distribution of multi-node verification results are dynamically displayed through a visual interface.

5. The algorithmic credit scoring method based on decentralized continuous performance authentication according to claim 4, characterized in that, The dynamic threshold δ is calculated as follows: Where: β is the algorithm robustness coefficient (0 < β ≤ 1); σ is the standard deviation of historical validation results; μ is the mean of historical validation results; I o This is the initial verification value.

6. The algorithmic credit scoring method based on decentralized continuous performance authentication according to claim 4, characterized in that, The consensus confirmation also includes: If performance metrics exhibit abnormal fluctuations during a single round of verification: Automatically trigger cross-regional node verification; The hash of the retest result is compared with that of the original result; If there is inconsistency, freeze the algorithm score and initiate a manual audit process.

7. The algorithmic credit scoring method based on decentralized continuous performance authentication according to claim 4, characterized in that, The end-to-end tracing includes: To query any verification event using a blockchain explorer, perform the following steps: Associated resource package Merkle tree path; Participating node reputation weight distribution; Score adjustment decision tree log; Supports multi-dimensional filtering and verification of records by time, geographic location, and dataset version.

8. An electronic device comprising a processor and a memory, characterized in that: The memory stores executable instructions; when the processor executes the instructions, it implements the steps of the method as described in any one of claims 4-7, and: The device operates as a consensus node in a decentralized verification network; Encrypted fragmentation of local storage resource layer data; The verification script is executed in isolation from the host operating system.

Images