Multi-dimensional dynamic personal value assessment and management system and method based on privacy computing

Abstract

This invention belongs to the fields of computer data processing, artificial intelligence, and blockchain technology, specifically relating to a multi-dimensional dynamic personal value assessment and management system and method based on privacy computing. Addressing the problems of existing technologies in personal value assessment, such as single dimensions, data silos, static models, and lack of credibility in assessment results, this invention proposes a multi-dimensional dynamic personal value assessment and management system and method based on privacy computing. This method constructs a secure data aggregation layer, utilizing secure multi-party computation or federated learning to fuse multi-source heterogeneous data without data leaving its domain; it comprehensively assesses personal value based on a multi-dimensional indicator system and machine learning models; it achieves continuous evolution of value scores through a dynamic update mechanism; and it uses blockchain to record the assessment process and results, ensuring data immutability and full traceability. This invention enables multi-dimensional quantification, dynamic assessment, and credible management of personal value, and can be widely applied in fields such as credit assessment, talent evaluation, financial risk control, and digital asset management.

Description

Technical Field

[0001] This invention belongs to the fields of computer data processing, artificial intelligence and blockchain technology, and specifically relates to a method and system for managing personal value data that supports a multi-dimensional evaluation model. Background Technology

[0002] With the acceleration of digitalization, data has become a core production factor, making it crucial to ensure its authenticity, integrity, and credibility. Existing personal value assessment systems generally suffer from limited assessment dimensions, severe data silos, a lack of dynamic model updates, opaque assessment processes, and a lack of reliable data storage. While blockchain technology can achieve data storage, it lacks deep integration with assessment models. Blockchain technology, with its distributed, immutable, and traceable characteristics, offers a revolutionary solution for building a decentralized, reliable data storage system. Numerous studies and practices both domestically and internationally have focused on blockchain-based data storage platforms; the following section introduces and analyzes the main categories of existing technologies.

[0003] (I) General-purpose blockchain evidence storage platform These platforms are typically built on public or consortium blockchains, providing standardized data hashing and on-chain notarization services. Users generate unique digital fingerprints (hash values) for their original data using hash algorithms (such as SHA-256) and write these hash values ​​to the blockchain. Due to the uniqueness of hash values ​​and the immutability of the blockchain, any modification to the original data will result in a change in its hash value, which can then be verified and discovered. For example, Ethereum-based notarization DApps (decentralized applications) allow users to pay gas fees to store file hashes in smart contracts. Domestically, there are also notarization services based on consortium blockchains such as FISCOBCOS and AntChain, providing enterprises with relatively controllable notarization environments. However, these general-purpose platforms have significant limitations: 1) Limited notarization content: They only store data hashes, lacking rich descriptions of the data's structure, attributes, and business logic, resulting in low "information density" and difficulty in supporting complex business scenarios. 2) Weak Off-Chain Connection: After the hash value is uploaded to the blockchain, the connection between it and the original off-chain data depends on the user's maintenance. This poses a risk of "decoupling between on-chain evidence and off-chain data," meaning that even if the on-chain hash is correct, the corresponding off-chain file may be replaced or lost. 3) Lack of Value Dimension: A dynamic and verifiable mapping relationship between the evidence and the rights it represents (such as property rights, copyrights, and creditor's rights) has not been established, meaning that "evidence" and "value" are disconnected. 4) Poor Verification Experience: Verification requires users to manually provide the original file to calculate the hash for comparison, which is unfriendly to non-professional users and makes it impossible to conveniently verify the validity of the evidence (such as whether it has expired or been revoked) on-chain.

[0004] (II) Enhanced Evidence Preservation Platform for Specific Scenarios To overcome the shortcomings of general-purpose platforms, some solutions have been enhanced for specific domains. For example, in the field of judicial evidence preservation, platforms enhance the legal validity of evidence preservation by connecting to the blockchain with timestamp service agencies and authoritative judicial appraisal centers. Some solutions standardize the evidence preservation process, defining a data structure that includes metadata such as time, subject, and operation type, and uploading it to the blockchain. In copyright protection scenarios, platforms may integrate technologies such as digital watermarking and feature extraction, and connect with copyright registration agencies. In supply chain traceability, logistics node information is uploaded to the blockchain by assigning RFID or QR codes to physical goods. These solutions improve practicality in specific domains, but the following problems still exist: 1) High system coupling and poor scalability: Deeply bound to specific business logic and external institutions, it is difficult to quickly adapt to the needs of other industries. 2) Inconsistent value identification: Different scenarios use different identifiers (such as order numbers, copyright registration numbers, and logistics tracking numbers), lacking a unified "value anchor" that crosses domains and connects the digital and physical worlds, making it difficult for evidence preservation data from different systems to interconnect and forming "value silos." 3) Balancing evidence granularity and cost: Fine-grained end-to-end evidence storage generates massive amounts of data, resulting in high on-chain costs; coarse-grained evidence storage may lose key details, affecting traceability effectiveness. There is a lack of a flexible mechanism to dynamically manage the relationship between evidence granularity and value.

[0005] (III) Evidence storage technology combining Internet of Things and identifier resolution The development of the Industrial Internet and the Internet of Things (IoT) has spurred the demand for directly storing physical equipment data on the blockchain. Related technologies typically combine identifier resolution systems (such as Handle, OID, and Ecode) to assign unique digital identities to physical objects and periodically store their status data on the blockchain. For example, some industrial blockchain platforms store hashed sensor readings and operating condition data of equipment on the blockchain for quality traceability or equipment health management. The challenges of this type of technology are: 1) Separation of identification and storage: The identification system is mainly responsible for object identification and resolution. It is often loosely coupled with the blockchain storage system or only interfaces, resulting in insufficient correlation. Furthermore, the identifier itself does not possess the ability to protect the integrity of the stored content. 2) Difficulty in data value screening: The IoT generates continuous, massive amounts of time-series data, and not all data has equal storage value. Existing solutions lack a data screening and aggregation storage mechanism based on "value density," which can easily lead to wasted storage resources or the omission of critical data. 3) Lack of dynamic value association: The state, ownership, and context (such as location and compliance status) of physical objects are dynamic and changing. Existing static identification and periodic evidence storage cannot reflect this dynamic value relationship in a real-time and structured manner.

[0006] (iv) Asset preservation based on "Token" or NFT In recent years, non-fungible token (NFT) technology has provided a new approach to the confirmation and notarization of digital assets. NFTs create unique digital certificates on the blockchain through smart contracts, representing ownership of artworks, collectibles, virtual real estate, and more. Some solutions attempt to extend the NFT concept to the notarization of physical assets or broad data, creating a corresponding NFT for a piece of data or a physical object. The advantage of this approach is that it achieves "assetization" and "tradability" of the notarized evidence. However, its limitations are equally apparent: 1) Emphasis on ownership, neglecting process and state: The core of an NFT is an ownership certificate, which is insufficient for business notarization requiring recording complex processes, multiple state transitions, and multi-party collaboration (such as contract signing processes and supply chain event sequences). 2) Vulnerable association with off-chain entities: The metadata of NFTs (such as URLs pointing to physical objects or digital files) is usually stored on off-chain centralized servers or decentralized storage such as IPFS. Its security, durability, and immutability are far lower than the on-chain portion, posing a risk of "metadata tampering" or "link failure," causing the NFT to become decoupled from the asset it represents. 3) Insufficient standardization and interoperability: The metadata formats, rights definitions and verification methods of NFTs issued by different platforms are different, making it difficult to be widely understood and accepted in different evidence storage and verification scenarios or cross-chain environments.

[0007] (v) Cross-chain evidence storage and verification technology With the development of multi-chain ecosystems, evidence storage data may need to be reliably transmitted and verified between different blockchain networks. Existing cross-chain technologies, such as relay chains, sidechains, and hash time locks, mainly focus on cross-chain asset transfers. When applied to evidence storage scenarios, they face the following challenges: 1) Semantic interoperability: The evidence storage data structures and business meanings differ across chains. How can we ensure that the verifier can correctly understand and verify the data after cross-chain transmission? 2) Lightweight verification: Full-node synchronous verification is too costly, requiring efficient lightweight verification solutions, such as those based on Merkle proofs or zero-knowledge proofs. However, existing evidence storage platforms generally do not natively integrate such cross-chain verification-friendly data structures.

[0008] As existing blockchain-based evidence storage technologies evolve towards a broader and deeper value internet, the core challenge they face is constructing a unified, strongly correlated, and dynamically scalable "data-evidence-value" mapping system. Specifically, this manifests in the following ways: 1) A universal "value identifier" is needed as an anchor point, capable of spanning the digital and physical worlds and compatible with various asset forms; 2) A strong, immutable, and efficiently verifiable binding relationship needs to be established between this identifier and the on-chain evidence storage content, preventing disconnect between on-chain and off-chain processes; 3) Flexible evidence storage and structured organization of rich attributes, states, and historical events surrounding this value identifier are needed to describe complex value transformation processes; 4) An efficient verification mechanism needs to be designed so that any relevant party can easily verify the authenticity, completeness, and validity of the declared value attributes of the evidence storage. Existing solutions either focus on single-point technologies for evidence storage (such as hash-based on-chain storage), or delve into vertical scenarios but have closed systems, or introduce new concepts (such as NFTs) but lack sufficient guarantees for relevance and integrity. None of these solutions systematically address the aforementioned challenges, thus hindering the large-scale, cross-domain application of blockchain-based trusted evidence storage technology in empowering the real economy and building a trusted digital society. Summary of the Invention

[0009] The technical problems that need to be solved in this invention are: This addresses the issue of fragile and easily decoupled association between on-chain hashes and off-chain original data in traditional blockchain notarization, ensuring the integrity and non-repudiation of notarized data throughout its entire lifecycle. This addresses the lack of unified, machine-readable value identifiers and structured descriptions for multi-source heterogeneous data (such as digital files, IoT streaming data, and business process status) during evidence preservation, enabling strong binding and parsing of evidence preservation data with the value of its physical or digital assets. To address the problem that static evidence storage cannot effectively reflect dynamic business processes and changes in asset status, a flexible mechanism is provided that supports value status updates, event tracing, and associated evidence storage. To address the challenges of cross-institutional and cross-system data trust and verification, a lightweight and efficient cross-domain trusted verification capability is built through standardized value code parsing and verification interfaces.

[0010] This invention aims to overcome the technical limitations of existing personal value assessment and management systems, such as single assessment dimensions, data silos, static and rigid models, high privacy and security risks, and difficulty in quantifying and circulating value. Specifically, it addresses the following technical problems: How can we construct a comprehensive and multi-dimensional assessment system to depict an individual's overall value, avoiding the one-sidedness of existing technologies that focus solely on financial credit or a single skill? How can we securely and compliantly aggregate and process heterogeneous personal data scattered across different institutions and platforms, breaking down data silos, while ensuring user data privacy and sovereignty? How can we design a dynamic and evolvable value assessment model that adapts to individual growth and changes in social evaluation standards, overcoming the rigidity of traditional static models? How can we ensure the transparency, interpretability, and immutability and traceability of the value assessment process, establishing credibility for the assessment? How can we transform abstract personal value data into standardized, verifiable, and potentially liquid digital assets, promoting the recognition and application of personal value in a wider range of scenarios?

[0011] Technical solution To address the aforementioned technical challenges, this invention provides a personal value data management system and method based on a multidimensional assessment model. This solution constructs a hierarchical and collaborative technical architecture, integrates multi-source data through privacy-preserving computation, utilizes a dynamic machine learning model for multidimensional value assessment, and leverages blockchain technology to achieve process notarization and result anchoring, ultimately outputting a verifiable and manageable digital profile of personal value and asset certificates.

[0012] A method for managing personal value data based on a multidimensional assessment model, characterized by the following steps: Step 1: Construct a secure multi-source data aggregation layer based on privacy-preserving computation. Addressing the dispersed and sensitive nature of personal data, a secure data access framework is designed. Data source providers (such as financial institutions, educational platforms, employers, and non-profit organizations) deploy data proxy nodes locally. When value assessment is required, secure multi-party computation (MPC) or federated learning (FL) protocols are used to perform joint statistics and feature engineering on the encrypted or anonymized original features, generating intermediate feature vectors for assessment, while the original data remains locally at the data source provider, achieving "data not leaving the domain." This layer abstracts a unified secure data access interface, supporting standardized access and privacy processing of structured (e.g., transaction records), semi-structured (e.g., project descriptions), and unstructured (e.g., text, images) data.

[0013] There are 𝑛 data source nodes, and the data in each node is: Splitting via a secret sharing mechanism: Aggregation via aggregation functions: Step 2: Design and train a multidimensional dynamic personal value assessment model. Define a multidimensional indicator system for personal value, including at least four categories: financial capital (repayment ability, asset status), human capital (educational background, skills certification, professional achievements), social capital (social network influence, reputation endorsement), and contribution capital (public welfare record, knowledge sharing, innovative achievements).

[0014] Constructing user feature vectors: Each dimension includes several quantifiable sub-indicators. An initial evaluation model is constructed using machine learning methods (such as gradient boosting trees and deep neural networks). Model training relies on the secure convergence features output from step 1, and uses anonymized, cross-domain success cases or expert-annotated data as supervision signals. The model incorporates a built-in feedback and online learning mechanism, periodically performing automated fine-tuning or triggering retraining based on new user data input, application feedback from evaluation results, and changes in macroeconomic market indicators, ensuring the model's dynamic adaptability and timeliness.

[0015] definition: Attention weighting: Dynamically updated: or: Step 3: Implement blockchain-based assessment process notarization and result anchoring. Blockchain is introduced as a trusted notarization and arbitration layer. Key assessment actions generate corresponding hash notarizations on the blockchain, including: a) authorization records and privacy computation task IDs for each data call request; b) feature vector hashes and model version numbers used for model inference; c) the final multi-dimensional value score and sub-item results. These notarizations constitute an immutable assessment audit trail. Furthermore, the final comprehensive value index (e.g., a standardized "value code"), along with timestamps and metadata, is minted on the blockchain in the form of digital credentials (e.g., verifiable credentials, VC), generating a uniquely identified "personal value digital asset certificate." This certificate can be held by the user and presented to a verifier when needed, allowing the verifier to verify the authenticity and status of the certificate on the blockchain.

[0016] Blockchain-based evidence storage: Value certificate: Step 4: Implement visualized management and application interface output of valuable data. Provide users with a personal data dashboard, visually displaying their multi-dimensional value scores, historical trends, contribution analysis of each dimension, and an overview of data sources. Simultaneously, the system provides standardized API interfaces, securely outputting verified value scores or asset certificates to third-party application scenarios, such as credit approval by financial institutions, talent screening on recruitment platforms, and rights redemption in partner communities, with user authorization. All data output adheres to the principle of minimum necessity and a dynamic authorization mechanism. Map the comprehensive score and multi-dimensional scores to the user dashboard or API output: Trend: Historical Change Curve Contribution: Contribution percentage in each dimension Support the principle of minimum necessity and dynamic authorization mechanism Furthermore, in step 1, the secure data aggregation layer adopts a threshold collaborative computing protocol based on Shamir secret sharing. Assume there is... n Each data source node participates in the computation, with a threshold value of t (t ≤ n). Each node outputs its local feature vector. v_i pass Shamir The algorithm divides the data into n shares and distributes them to the other nodes. Each node collects at least... t The aggregate feature function F(v_1, ..., v_n) can be recovered through Lagrange interpolation (such as summation or mean), but the original vector of any individual node cannot be obtained. v_i This method, while ensuring data privacy, tolerates the offline status of some nodes (no more than nt).

[0017] Further, in step 2, the multi-dimensional dynamic evaluation model employs an attention-weighted multi-task learning network structure. The model input consists of feature groups G_f, G_h, G_s, and G_c from different dimensions. First, each feature group undergoes feature extraction and dimensionality reduction through an independent sub-network (fully connected layer) to obtain preliminary dimensional embedding vectors e_f, e_h, e_s, and e_c. Then, a shared attention module learns the dynamic weights α_f, α_h, α_s, and α_c for each dimension with respect to the current evaluation target (e.g., "comprehensive credit" and "career potential"), satisfying Σα = 1. Finally, the weighted and aggregated vectors are input to the downstream multi-task output layer, simultaneously generating a comprehensive score S_{total} and sub-scores S_f, S_h, S_s, and S_c for each dimension. The loss function is a weighted sum of the comprehensive score task loss and the sub-score task losses, and is continuously optimized using feedback data.

[0018] Furthermore, in step 3, the blockchain adopts a hybrid architecture of "main chain-side chain". The main chain (such as a public chain or a large consortium chain) is responsible for storing the metadata hash and global state of the core credentials, ensuring maximum immutability and credibility. The side chain (or application-specific chain) is responsible for handling the high-frequency evaluation process and reducing the load on the main chain. The side chain and the main chain conduct periodic settlements through lightweight relays or state anchoring. The "personal value digital asset certificate" is implemented in accordance with ERC-1155 or similar multi-token standards, giving it uniqueness, portability, and rich programmability, enabling it to associate complex metadata and circulate and verify in different decentralized applications (DApps).

[0019] Beneficial effects The personal value data management system and method based on a multidimensional evaluation model, as described in this invention, have the following beneficial effects: The comprehensiveness and objectivity of the assessment have been significantly improved: By integrating multi-dimensional data such as finance, human resources, society, and contribution, and using machine learning models for non-linear synthesis, the one-sidedness of traditional single-dimensional assessments has been overcome. This allows for a more comprehensive and objective reflection of an individual's overall value, providing richer references for various decision-making processes.

[0020] Effectively breaking down data silos and ensuring privacy and security: Employing privacy-preserving computing technologies (such as secure multi-party computation and federated learning) to achieve "data usable but not visible" enables secure joint value assessment without the need to centralize raw data. This promotes the integrated utilization of data value and fundamentally protects user data privacy and the data sovereignty of the source party, in line with increasingly stringent data regulatory requirements.

[0021] Achieving dynamic and personalized assessment models: The models have online learning and feedback optimization mechanisms, which can automatically adjust as time goes by and new data is added, adapting to changes in individual growth and social evaluation systems, making assessment results more timely and personalized, and overcoming the lag of static models.

[0022] Establishing a credible and traceable system for the assessment process: By hashing key steps onto the blockchain, a complete and tamper-proof assessment audit trail is formed, greatly enhancing the transparency and credibility of the assessment process. Users and regulators can trace the scoring basis, and questionable results can be verified and arbitrated, improving the fairness and credibility of the system.

[0023] Promoting the assetization and circulation of personal value data: Evaluation results are forged into standardized, on-chain verifiable digital asset certificates, transforming abstract personal value into concrete, self-manageable digital assets. This provides an efficient and reliable technological carrier for the secure transfer, verification, and application of personal value across platforms and scenarios (such as credit lending, talent certification, and rights exchange), contributing to the construction of a more vibrant value exchange ecosystem.

[0024] Enhancing user experience and data sovereignty awareness: Providing users with an intuitive data dashboard allows them to clearly understand the composition and changes of their own value. Through self-authorized management of data use and the presentation of value credentials, it strengthens users' sense of control and sovereignty over their digital assets, aligning with the development trends of Web3.0 and self-sovereign identity. Detailed Implementation

[0025] This invention aims to address the problems of single-dimensional personal value assessment, data silos, static models, high privacy risks, and difficulty in quantifying and circulating value in existing technologies. It provides a personal value data management system and method based on a multi-dimensional assessment model. To clearly illustrate the technical solution of this invention, the specific implementation methods will be described in detail below, combining system architecture and data circulation processes. The core of this invention lies in constructing a layered, collaborative technical system that securely integrates multi-source data, performs comprehensive assessment through a dynamic intelligent model, and utilizes blockchain technology to ensure process credibility and result verifiability. Ultimately, it outputs a personal digital value profile and asset certificate with practical application value.

[0026] The system's overall architecture can be divided into four logical layers: a data security aggregation layer, an intelligent evaluation engine layer, a blockchain notarization and anchoring layer, and an application interaction and service layer. The data security aggregation layer, located at the bottom layer, is responsible for securely connecting with various external data sources; its feature vector is denoted as: n: Total number of data source nodes d: Number of feature dimensions These data sources are highly heterogeneous and may include credit transaction records from commercial banks, consumption records from third-party payment platforms, academic credentials from the Ministry of Education's Academic Credentials Verification System, professional qualification certification databases from human resources and social security departments, work experience and skill tags from large recruitment platforms, interaction and content contribution data from social media platforms, volunteer service records from public welfare organizations, and patent or copyright registration information from intellectual property agencies. To achieve collaborative computing under the premise of "data not leaving the domain and privacy not being leaked," this layer requires the deployment of a lightweight "data collaboration proxy gateway" in the local network environment of important data sources. This gateway software module has dual functions: first, internally, it extracts and preprocesses the required feature data from local business databases or data warehouses on demand through secure interfaces (such as authenticated and encrypted APIs). Preprocessing includes format standardization, anonymization (such as converting precise amounts into range values), and necessary feature encoding; second, externally, it participates in collaborative computing tasks in accordance with a unified privacy computing protocol. When the central scheduling system (initiated by the evaluation engine layer) initiates a personal value evaluation request, it generates a unique task identifier and notifies the proxy gateways of the relevant data sources through the task coordinator. After obtaining prior authorization from users for data use, which is time-limited and scope-bound (the authorization record itself needs to be stored on the blockchain for notarization), each gateway does not directly exchange raw data. Instead, based on a predetermined computation graph (e.g., calculating the user's average monthly stable income over the past 24 months, consumption diversity index, weight value corresponding to the highest education level, time decay coefficient of the most recent skills certification, effective connectivity of social networks, etc.), it executes a portion of the computational subtasks locally. For aggregation computations requiring data from multiple data sources, a secure multi-party computation or federated learning approach is adopted. Shamir secret sharing is used for splitting. Aggregate functions: Each node only sends encrypted / sharding information. Any 𝑡 shards can be used to recover the aggregation result through Lagrange interpolation, protecting the privacy of the original data while tolerating 𝑛-𝑡 nodes being offline.

[0027] For example, to estimate a user's overall debt level, it might be necessary to securely sum data across multiple financial institutions. In this case, each financial institution's proxy gateway uses homomorphic encryption or secret sharing technology to process its local debt amount, generating encrypted intermediate values ​​or data fragments. Only this information, which cannot be reversed to the original data, is sent to designated secure computing nodes or directly interacts with each other using secure protocols. Finally, the coordinating party or a threshold number of participating parties collaborate to decrypt the aggregated result (total debt), and no single participating party knows the specific data of the others. This layer encapsulates the complexity of interfacing with different data source systems by defining a standard adapter interface and provides a unified, privacy-preserving feature data service to the upper-layer evaluation engine. It uniformly generates intermediate feature vectors for different types of data (structured, semi-structured, and unstructured). Where 𝜙 is the standardization, desensitization, and feature encoding function.

[0028] The intelligent assessment engine layer is the core computing and decision-making hub of the system. It receives standardized feature vectors from the data security aggregation layer, which have undergone preliminary cleaning and privacy processing. The primary task of this layer is to run the "multi-dimensional dynamic personal value assessment model." This model is not a single algorithm, but a complex, configurable model pipeline or integrated system. Its implementation begins with a carefully designed "value dimension indicator system." This system is the blueprint for model construction, usually defined jointly by domain experts and data scientists, and reserves the possibility of iterative updates through a democratic community governance mechanism. The indicator system adopts a tree structure, with the root node being "personal comprehensive value," under which several primary dimensions are expanded, such as "credit performance dimension," "capital accumulation dimension," "ability and skills dimension," "social influence dimension," and "contribution creation dimension." Each primary dimension is further subdivided into observable and quantifiable secondary and even tertiary indicators. For example, the "ability and skills dimension" may include secondary indicators such as "educational background," "professional certification," "project experience," and "mastery of the technology stack"; "educational background" can be further divided into tertiary indicators such as "academic level," "institutional reputation," "major relevance," and "academic performance." Each underlying metric is associated with one or more data sources (corresponding to features in the data aggregation layer) and a quantization transformation rule (e.g., mapping educational level to numerical scores). Embedding vectors are extracted through sub-networks. * indicates the input from the previous text.

[0029] The evaluation engine incorporates various machine learning model components to handle different types of evaluation tasks. For predicting comprehensive scores, gradient boosting decision trees (such as XGBoost and LightGBM) or deep neural networks may be employed. These models are trained on massive amounts of anonymized training samples (the labels of which may come from historical credit performance, high-end job hiring decisions, selection of important project collaborations, and other comprehensive success indicators), learning the complex non-linear relationships between various dimensional features and the final value potential. The model training process itself also emphasizes interpretability, using attention mechanisms for weighting. Weighted aggregation yields a comprehensive embedding: Overall score and sub-dimension scores: Loss function: It can simultaneously generate an analysis report on the contribution of each feature to the current score, which users can review, thus solving the "black box" problem.

[0030] More importantly, this assessment engine is "dynamic." It continuously monitors several types of signals: first, the characteristic changes brought about by the influx of new user data; second, the actual feedback after the assessment results are used by downstream applications (such as lending institutions) (e.g., whether the subsequent repayment performance of loans issued based on the system's score matches the score prediction); and third, the changing trends of macroeconomic and social talent evaluation standards (through access to external macroeconomic data sources). The system periodically (e.g., monthly) or triggered (when feedback deviation exceeds a threshold) initiates the incremental learning or fine-tuning process of the model, dynamically updating the formula: and time decay: The engine automatically adjusts model parameters and even explores better model structures through automated machine learning (AutoML) to ensure that evaluation standards are up-to-date. Furthermore, it includes a "contextualized evaluation" module, which allows for dynamic adjustment of indicator weights or activation of different sub-models based on different application scenarios (such as job seeking, lending, and social matching), thereby outputting more scenario-specific value analysis reports.

[0031] The blockchain notarization and anchoring layer provides a trust foundation for the entire system. This layer does not directly process massive amounts of raw data, but focuses on recording hash digests and state changes of key actions, enabling traceable and tamper-proof audit trails. The system adopts a layered blockchain design to adapt to different performance and trust requirements. The bottom layer uses a widely trusted "main chain" (which may be a permissioned consortium blockchain maintained by multiple authoritative institutions; or, when conditions are ripe in the future, a specific public chain) as the final trust anchor. The upper layer can build multiple high-performance "application sidechains" or "subchains," each responsible for handling high-frequency notarization transactions in different regions or of different types. Any operation that substantially affects personal data must generate a corresponding notarization record. The specific process is as follows: When a user registers with the system for the first time or authorizes a new assessment, the hash value of this authorization action (including the scope of authorization, validity period, and target data source) is recorded on the chain, forming "authorization notarization."

[0032] After the data security aggregation layer executes the privacy computation task, the proof of the task's success, the hash of the aggregated result of the feature vectors used (representing the data fingerprint used in this computation), and the identifier of the data source participating in the computation together constitute the "computation process evidence" on the blockchain. After the evaluation engine layer generates the scoring result, it packages the model version number on which the evaluation is based, the input feature hash, and the output scoring result (including the comprehensive score and scores for each dimension) together to generate an "evaluation result evidence" and submits it to the blockchain. These evidences are like unforgeable "postmarks," interlocking with each other, allowing anyone to trace the complete origin of a score. Finally, the system encapsulates the periodic comprehensive evaluation results (e.g., a comprehensive value index updated quarterly) into a standardized digital certificate—a "personal value digital asset certificate."

[0033] The credential is constructed in accordance with the W3C Verifiable Credentials (VC) standard or similar open standards, and includes a holder identifier (decentralized identifier DID), issuer identifier (system operator), credential content (score, validity period), digital signature, and a query index pointing to the relevant on-chain evidence. The credential's metadata (such as credential ID, status, and issuance time) hash is anchored to the main chain, while the complete data of the credential itself can be stored in a user-controlled digital wallet or decentralized storage network. Users have complete control over the credential and can choose when and to whom to present it. After receiving the credential, the verifier can verify its authenticity and validity (whether it has been revoked or expired) by querying the on-chain anchored hash, without directly connecting to a central database, thus achieving decentralized trust verification. The sidechain and the main chain are synchronized periodically (e.g., daily) through lightweight state channels or relays, ensuring that evidence-gathering activities on the sidechain can ultimately find tamper-proof evidence on the main chain.

[0034] The application interaction and service layer serves as the interface between the system and end users and third-party applications. For individual users, the system provides a fully functional personal data and value management center, typically presented as a mobile application or web dashboard.

[0035] Users can: view and manage all their associated data source authorizations in one place; view their multidimensional value score, historical trends, comparative analysis with peers or industry peers, and interpretation of key factors influencing the score in an intuitive dashboard and chart format; review every important data usage and evaluation record stored on the blockchain, enhancing transparency and control; and manage their "personal value digital asset certificate," performing actions such as presentation, temporary authorization, or revocation. For third-party application providers such as enterprises or institutions, the system provides a set of standard, secure, and open API interfaces. These interfaces allow application providers to initiate evaluation requests or query specific value information (not necessarily the original score, but judgments that meet specific scenarios, such as "whether they meet the credit threshold for applying for this product") after obtaining explicit authorization from the user (this authorization operation must be recorded on the blockchain). All external data services are subject to strict access control, traffic monitoring, and auditing. In addition, the system can support more advanced application models, such as the prototype of a "value data market": users can selectively list their value scores or analysis reports (still anonymized or generalized) on the market for research institutions or enterprises in need to purchase. The revenue is distributed among users, data providers and the platform according to preset smart contract rules, thereby exploring the direct monetization path of personal data value.

[0036] The entire system begins with user registration and identity verification. After successful authentication, the system generates a globally unique decentralized identifier (DID) and a pair of asymmetric encryption keys for the user, with the private key securely kept by the user. The user then accesses the data authorization management interface, where they can browse various data sources supported by the system. When a user wants to add a data source (e.g., their credit card statement), the system guides them through standard protocols such as OAuth to the official authorization page of that data source provider, completing secure account association and data scope authorization. The hash value of the final result of this authorization process (an authorization token or statement signed by the data source provider) is associated with the user's DID and recorded on the blockchain, serving as the legal basis for all subsequent data access. Users can also revoke any authorization at any time in their personal center; this revocation is also recorded on the blockchain, and the relevant data source will immediately cease to be accessed.

[0037] When a user initiates a value assessment, or a third-party application (with user authorization) requests to assess the user, the system's task scheduler begins operation. The scheduler first verifies the legitimacy of the request (checking records on the authorization chain), and then selects the appropriate model pipeline configuration from the assessment model library based on the assessment objective (default comprehensive assessment or specific scenario assessment). This configuration explicitly lists all features required for this assessment and their data sources. Next, the scheduler issues the computation task to the task coordinator in the data security aggregation layer. Based on the feature list, the task coordinator identifies the data source proxy gateways that need to participate in this computation and sends them computation instructions and necessary parameters (e.g., session ID for secure multi-party computation, current-round model parameters for federated learning). Each proxy gateway performs computation locally, coordinating through a privacy computation protocol, and ultimately returns the computed feature values ​​(in plaintext or in a form still encrypted but usable for subsequent computations) to the assessment engine. The assessment engine aggregates all features, inputs them into the selected assessment model, runs inference computation, and generates a draft report containing a comprehensive score, dimensional scores, and contribution analysis. Before the final report is generated, the engine submits key information from this evaluation (model version, input feature hash, output results) to the blockchain notarization layer and receives a receipt containing the notarized transaction ID. Finally, the system presents the complete evaluation report, interpretability analysis, and blockchain notarization receipt (including a transaction query link) to the user or authorized applicant. Simultaneously, if this evaluation generates new "personal value digital asset certificates" (e.g., at quarterly update times), the system invokes the blockchain layer's certificate issuance service to mint and distribute new certificates to the user.

[0038] During system operation, a dynamic learning mechanism runs continuously in the background. A separate "model maintenance module" collects follow-up feedback data from each evaluation (obtained with user authorization and under compliance conditions, such as whether the user successfully obtained a loan based on a high score and made timely repayments), and calculates the deviation between the predicted value and the actual result. When the deviation statistic exceeds a preset threshold, or when the predetermined model retraining cycle is reached, this module automatically prepares a new training task. The training task is also conducted in a privacy-preserving manner, coordinating various data sources to use historical data for federated learning training and updating model parameters. After rigorous offline testing and A / B testing, the new model is gradually rolled out to replace the old model. All change history of the model version, including training data overview, performance metrics, deployment time, etc., is stored on the blockchain to ensure the transparency and auditability of model evolution.

[0039] To handle large-scale concurrent requests and ensure high system availability, all components in the aforementioned layers are designed as horizontally scalable microservice architectures. The proxy gateway in the data security aggregation layer can be deployed in a distributed manner; the evaluation engine can launch multiple model inference instances, with requests distributed by a load balancer; and the blockchain layer can enhance transaction processing capabilities through sidechain sharding technology. All communication between services uses TLS encryption, and internal access follows the principle of least privilege. The system also has a dedicated regulatory interface, allowing compliance and auditing agencies to query and verify on-chain evidence, fulfilling necessary regulatory obligations, provided that legal procedures are followed.

[0040] This invention constructs a complete closed loop from secure data access to intelligent comprehensive evaluation, and then to trusted storage and value output. It not only provides a more scientific tool for measuring personal value, but more importantly, it establishes a trust system based on a technological architecture, enabling personal value data to be securely generated, transferred, and applied while protecting privacy and clarifying ownership. This provides a solid technological infrastructure for releasing personal value in the digital society. The entire system embodies the design philosophy of "user-led, privacy-first, algorithmic transparency, and value linkage," representing a systematic innovation to the existing decentralized, opaque, and static personal data management and evaluation models.

Claims

1. A multi-dimensional dynamic personal value assessment and management system and method based on privacy computing, characterized in that... A data security aggregation layer is constructed, which, through secure multi-party computation or federated learning, jointly processes personal data from multiple data sources without transmitting the original data, generating multi-dimensional feature vectors. Based on these multi-dimensional feature vectors, a multi-dimensional personal value assessment model is built to quantitatively evaluate personal value, obtaining a comprehensive value score and multiple dimension sub-scores. The comprehensive value score is dynamically updated based on time series and changes in new data. Key data, model parameters, and assessment results from the assessment process are hashed and stored on the blockchain to achieve trusted evidence storage of the assessment process. A digital certificate of personal value is generated based on the assessment results, and query and verification interfaces are provided.

2. The data security aggregation layer employs a threshold computing mechanism based on secret sharing to achieve multi-party data collaborative computation. The multi-source data includes at least one of behavioral data, asset data, social data, capability data, and contribution data. The multi-dimensional feature vector is represented as: .

3. The multidimensional evaluation model includes a machine learning model, a graph structure model, and a time-dynamic model. The model calculates weights based on an attention mechanism, and its comprehensive value calculation formula is as follows: ,in: The dynamic update uses the following formula: ,in: .

4. The blockchain-stored evidence includes: The features include a feature vector hash, model version number, evaluation result, timestamp, and digital signature. The personal value digital certificate uses a verifiable certificate structure and supports on-chain verification. The system claims include a data security aggregation module, a feature processing module, a multi-dimensional evaluation model module, a dynamic update module, a blockchain notarization module, a value certificate generation module, and a query and visualization module.

Images