Human resource information sharing system and method based on big data

Abstract

The invention discloses a human resource information sharing system and method based on big data, and relates to the technical field of big data processing and information security, and the system comprises a trusted center node and a plurality of distributed local nodes; the local node is used for storing and processing original human resource data of the mechanism; the trusted center node is used for coordinating federal learning tasks, maintaining a global model and encrypting an index database; the system also comprises a dynamic portrait construction module which is deployed at each local node and is used for generating a personal dynamic skill portrait or an organization demand portrait according to the local data and the global model; a privacy calculation matching engine; on the premise that original data are not collected, a global intelligent model is constructed through federal learning, a real-time state is described by utilizing a dynamic portrait, safe matching of data availability and invisibility is realized by virtue of privacy computing technologies such as attribute-based encryption and the like, and finally an efficient, accurate and compliant human resource ecological collaboration platform is constructed.

Description

Technical Field

[0001] This invention belongs to the field of big data processing and information security technology, specifically relating to a human resources information sharing system and method based on big data. Background Technology

[0002] With the deepening development of the digital economy, the optimal allocation of human resources increasingly relies on data-driven precise matching. However, the current human resources market suffers from prominent data silos, with data systems from enterprises, recruitment platforms, and educational institutions not interconnected, making it difficult to form a comprehensive and dynamic talent insight; information is static and distorted, traditional resumes are outdated and cannot reflect the evolution of individual skills and changes in organizational needs in real time; finally, privacy and security risks are becoming increasingly serious, and centralized data storage and plaintext exchange models are prone to leakage of sensitive information, making it difficult to meet increasingly stringent data compliance requirements.

[0003] Existing technological approaches in the market, whether relying on traditional recruitment platforms that depend on centralized data management or decentralized blockchain solutions that emphasize user autonomy, are insufficient to fully meet the demands. While the former can leverage its accumulated resume database for keyword matching, it struggles to avoid issues such as data monopoly, privacy risks, and single points of failure. The latter, while strengthening data sovereignty through distributed mechanisms, has limitations in supporting complex intelligent matching and deep data analysis, and often comes with high performance overhead.

[0004] Therefore, there is an urgent need to build a human resources information sharing system and method that can achieve dynamic, accurate, secure sharing and intelligent matching of human resources information while protecting the data privacy and sovereignty of all parties.

[0005] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a human resources information sharing system and method based on big data. It can build a global intelligent model through federated learning without collecting raw data, use dynamic profiling to depict real-time status, and use privacy computing technologies such as attribute-based encryption to achieve secure matching of data that is available but not visible, ultimately building an efficient, accurate, and compliant human resources ecosystem collaboration platform.

[0007] To achieve the above objectives, a specific embodiment of the present invention provides the following technical solution: A big data-based human resources information sharing system includes: One trusted central node and multiple distributed local nodes; The local node is used to store and process the organization's raw human resources data; The trusted central node is used to coordinate federated learning tasks, maintain the global model, and maintain the encrypted index library. The system also includes: The dynamic profile building module is deployed on local nodes and is used to generate dynamic personal skill profiles or organizational need profiles based on local data and the global model. A privacy-preserving computation matching engine is deployed on and integrated into the trusted central node. It is used to encrypt and query profiles using attribute-based encryption technology and perform matching calculations in the ciphertext domain. The anonymous matching and authorization module is used to provide anonymous communication and a controllable information hierarchical decryption channel for both parties who are successfully matched.

[0008] In one or more embodiments of the present invention, the personal dynamic skill profile is generated by fusing static attributes and dynamic behavioral data. The dynamic behavioral data includes at least one of authorized code repository contribution records, learning behavior data from online education platforms, or collaboration data from project collaboration tools, and skill proficiency and activity are quantified through a time-series model.

[0009] In one or more embodiments of the present invention, the global model is updated through the following federated learning process: the trusted central node distributes the current global model to each local node; each local node calculates model parameter updates or gradients based on local data; each local node uploads the encrypted updates or gradients to the trusted central node; the trusted central node performs secure aggregation to generate a new version of the global model and distributes it.

[0010] In one or more embodiments of the present invention, the secure aggregation employs homomorphic encryption or secure multi-party computation techniques to ensure that the trusted central node cannot know the original update content of any local node during the aggregation process.

[0011] In one or more embodiments of the present invention, the privacy-computing matching engine includes: The encryption unit is used to encrypt the image into a ciphertext index using an attribute-based encryption algorithm, and its decryption strategy is associated with the attributes of the image. The query processing unit is used to convert the queryer's search criteria into search tokens that are bound to their attributes; The ciphertext matching unit is used to calculate the matching degree between the search token and the ciphertext index without decryption.

[0012] In one or more embodiments of the present invention, the anonymous matching and authorization module is configured to: in the initial matching stage, only return anonymous candidate identifiers and desensitized matching scores to the querying party; after both parties reach a preliminary agreement through the anonymous channel, the queried party authorizes the decryption of more detailed information layers in its profile level by level.

[0013] In one or more embodiments of the present invention, a market trend analysis module is also included. The module generates a macro skills supply and demand heat map, salary bandwidth and liquidity analysis report based on the anonymized statistical information uploaded by each local node and processed with differential privacy, and feeds it back to each local node.

[0014] In one or more embodiments of the present invention, applied to a system including a trusted central node and multiple local nodes, the method includes: S1. Each local node uses local data to participate in federated learning coordinated by the trusted central node, and jointly trains and maintains a global dynamic profile model. S2. Each local node generates a dynamic profile of an individual or organization based on the global model and its own data, and uploads it to the encrypted index library of the trusted central node after encrypting it using attribute-based encryption technology. S3. When a local node initiates a query, the trusted central node receives its encrypted query request, performs privacy-preserving matching calculations in the encrypted index, and returns an anonymous result list that conforms to the attribute policy. S4. Through the anonymous communication mechanism provided by the system, a connection is established between the querying party and the candidates in the result list, and more information is exchanged step by step based on the candidates' active authorization.

[0015] In one or more embodiments of the present invention, step S1 specifically includes: S5. The trusted central node initializes the global model parameters and broadcasts them to all participating nodes; S6. Each participating node trains the model on its local dataset and calculates the parameter differences or gradients. S7. Each participating node encrypts the parameter differences or gradients before uploading them; S8. The trusted central node performs secure aggregation of encrypted information and updates global model parameters. S9. Repeat S6-S8 until the model converges.

[0016] In one or more embodiments of the present invention, in step S3, the privacy protection matching calculation specifically involves: converting the query request into a set of attribute policy-related search tokens, filtering out ciphertext indexes that meet the conditions through the ciphertext policy matching mechanism of the attribute-based encryption algorithm, and calculating the similarity of the portrait feature vectors in the encrypted vector space.

[0017] Compared with existing technologies, the big data-based human resources information sharing system and method of this invention, through a federated learning architecture, allows participating parties to collaboratively train advanced models without sharing raw data, fundamentally ensuring data privacy and compliance. By continuously accessing multiple dynamic data sources, it constructs personal and organizational profiles that evolve over time, greatly improving the timeliness and accuracy of information. Utilizing attribute-based encryption technology, it achieves encrypted matching between query conditions and profile data, ensuring that sensitive information is not leaked throughout the matching calculation process. Through anonymous matching and hierarchical authorization mechanisms, it builds a secure and controllable communication bridge for talents and employers while protecting privacy, and provides macro-market insights to empower decision-making for all parties. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the overall architecture of the human resources information sharing system in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the construction and updating of a personal dynamic skill profile in an embodiment of the present invention; Figure 3 This is a schematic diagram of the privacy-preserving matching process based on attribute-based encryption in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the overall steps of a human resources information sharing method in one embodiment of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions in this disclosure, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this disclosure.

[0021] like Figure 1 As shown, the system of the present invention adopts a hybrid architecture of federated data lake and trusted coordination center; Local nodes are entities with independent computing and storage capabilities, such as enterprise servers or personal client proxies. They deploy dynamic profile building modules and local databases. The local database stores raw data and can synchronize data from internal enterprise employees or candidates (resumes, performance evaluations, project experience), individual users (educational certificates, skills self-assessments), and data from third-party platforms explicitly authorized by the user). The core responsibility of local nodes is to process raw data locally, participate in federated learning, and generate encrypted dynamic profiles.

[0022] A Trusted Central Node is a neutral coordinator that does not belong to any single data provider. It is typically maintained jointly by alliance members or operated by a trusted third party. It comprises the following core components: The federated learning coordinator is responsible for organizing each round of federated training tasks, including the initialization and distribution of the global model, the collection and secure aggregation of encrypted updates.

[0023] An encrypted index repository stores ciphertext indexes uploaded by all local nodes, which have undergone attribute-based encryption. This index is an encrypted digest of the dynamic profile, containing ciphertext of key feature vectors used for matching, but not any directly identifiable personal information.

[0024] The privacy-computing matching engine includes: The encryption unit defines the encryption interface for attribute-based encryption technology. When a local node uploads an image, it needs to call the specification of this interface for encryption.

[0025] The query processing unit receives a search token sent by the querying party's local node, which has been processed with its own attribute private key.

[0026] The ciphertext matching unit performs the core ciphertext field calculation. It utilizes the properties of attribute-based encryption algorithms to directly process search tokens and ciphertext indices. It calculates whether the token attributes satisfy the access policy of the index and performs vector similarity calculation in the encrypted state for the indexes that satisfy the policy.

[0027] The anonymous matching and authorization module manages an anonymous identifier system and provides a communication channel for both parties. Simultaneously, this module maintains a hierarchical information decryption protocol, with data subjects authorizing decryption layer by layer.

[0028] The market trend analysis module periodically extracts aggregated statistical information processed with differential privacy technology from an encrypted index to generate visual reports. The aggregated statistical information includes the frequency of occurrence of a certain skill and its geographical distribution, but does not involve any individual.

[0029] Detailed description of the dynamic portrait construction process like Figure 2As shown, taking the generation of a personal dynamic skills profile as an example, this process is continuous and automated: Multi-source data collection: After obtaining explicit authorization from the user, the system periodically retrieves data from multiple data sources through a secure OAuth API. For example, it retrieves code commit records and repository star counts from GitHub every week; it retrieves course completion progress and grades from Coursera; and it retrieves task completion records from Jira, an internal project management tool for enterprises.

[0030] Feature extraction and standardization utilize pre-trained NLP models to analyze project descriptions, code comments, and course outlines to extract skill keywords; simultaneously, behavioral data is quantified into metrics such as the number of submissions in the past 90 days and the final course score.

[0031] Federated knowledge enhancement involves inputting the extracted raw features into a global skill recognition model trained through federated learning. Because this model learns from data from numerous enterprises, it can more accurately determine the proficiency level corresponding to a given skill.

[0032] For profile quantification and updates, the system maintains a time-series-based skill tag cloud. Each skill tag is associated with multiple dimensions: basic proficiency, recent activity level, and decay coefficient; the profile is updated every 24 hours or triggered by a specific event.

[0033] The updated profile is immediately encrypted using the local node's attribute-based encryption public key and a preset strategy to generate a ciphertext index, which is then uploaded to the encrypted index library of the trusted central node.

[0034] Detailed Explanation of Federated Learning Model Training The training of the global dynamic profiling model follows the federated learning process, as follows: Figure 4 As shown in the S1 stage: Initialization and distribution: The trusted central node initializes the model parameters and broadcasts them to all local nodes registered to participate in this training task.

[0035] Local computation involves each local node using its own private data to perform multiple batches of stochastic gradient descent locally, calculating the local model update Δ. or gradient∇ Before training, the data is thoroughly anonymized and desensitized locally.

[0036] Encrypted upload; to enhance security, the local node will update Δ Before being sent to the central node, it is encrypted using a homomorphic encryption algorithm to obtain... Even if the transmission process is intercepted or the central node is malicious, the original update content cannot be detected.

[0037] Secure aggregation: After receiving all encrypted updates, the trusted central node directly calculates the encryption using the additive homomorphism of homomorphic encryption. = Then, decrypt using its private key. , obtain the aggregated total update This process is known as secure multi-party aggregation, where the central node has never encountered the plaintext Δ. .

[0038] Model updates and distribution, with the central node following the formula Update the global model, where η is the learning rate, and add the new... The data is distributed to each node to begin the next round of training. This process iterates until the model converges.

[0039] Privacy-protected matching and anonymous matchmaking, such as Figure 3 and Figure 4 As shown in stages S3 and S4, a complete privacy-protected matching and matchmaking process is as follows: The query is initiated by Company A's HR department, who wants to find a researcher with 5 years of experience who is proficient in federated learning. The HR department inputs the query criteria into a local node. The query processing unit of the local node transforms these criteria into an attribute strategy and a feature query vector. The attribute strategy mainly includes company type and job role. The feature query vector is encoded by the global model using federated learning techniques to represent the text query.

[0040] For token generation and encrypted querying, the system uses Enterprise A's attribute private key to process the feature query vector and generate an encrypted search token. This token, along with the attribute policy, is sent to the trusted central node.

[0041] Ciphertext matching computation: The ciphertext matching unit of the privacy computation matching engine performs the following operations: The strategy involves filtering and traversing the encrypted index database. For each individual profile's encrypted index (CT), an attribute-based encryption algorithm is used to verify whether enterprise A's attributes satisfy the access policy embedded in the CT. If not, it is immediately excluded.

[0042] For similarity calculation, for CTs that pass the policy screening, the engine calculates the similarity between the encrypted query vector in Token TK and the encrypted feature vector in CT in the encrypted state; by designing a special homomorphic encryption algorithm to implement the inner product operation of the vectors, the encrypted similarity score E is obtained.

[0043] Upon receiving the results, the central node decrypts the encrypted scores, sorts them by score, and returns the anonymous IDs and corresponding de-identified scores of the top K candidates to the HR department of company A. At this stage, the HR department cannot see any plaintext information about the candidates.

[0044] Anonymous matchmaking and tiered authorization: HR can initiate a business card exchange request to an anonymous ID that is of interest. The system uses the anonymous matchmaking module to send an anonymous in-system message to the corresponding candidate, which could read: "A high-tech company is interested in your background; would you be willing to disclose your basic profile to them?"

[0045] Once the tiered information is unlocked, the candidate receives a notification and can view Company A's anonymized company profile. If interested, the candidate clicks "Agree." At this point, the anonymized matchmaking and authorization module is triggered, sharing the decryption key of the first-level encrypted text in the candidate profile with Company A, allowing Company A to view this information. If HR wishes to view more detailed project experience, a second request must be initiated, requiring secondary authorization from the candidate. The candidate retains ultimate control over their contact information and can refuse to provide it at any stage.

[0046] The market trend analysis module runs independently of the real-time matching process and is executed periodically. The data collection module requests statistical information that meets differential privacy requirements from all local nodes.

[0047] Aggregate analysis uses a central node to aggregate these noisy reports. Due to the characteristics of differential privacy, the aggregation results can closely approximate the true market distribution, while ensuring that no individual information can be inferred.

[0048] The report generation and feedback module generates insights such as the supply and demand ratio of blockchain talent in East China and the quarterly changes in the median salary of full-stack development engineers. These insights are then fed back to local nodes via dashboards or APIs to help companies develop recruitment strategies and individual learning paths.

[0049] It will be apparent to those skilled in the art that this disclosure is not limited to the details of the exemplary embodiments described above, and that this disclosure can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of this disclosure is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within this disclosure. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0050] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A human resources information sharing system based on big data, characterized in that, include: One trusted central node and multiple distributed local nodes; The local node is used to store and process the organization's raw human resources data; The trusted central node is used to coordinate federated learning tasks, maintain the global model, and maintain the encrypted index library. The system also includes: The dynamic profile building module is deployed on local nodes and is used to generate dynamic personal skill profiles or organizational need profiles based on local data and the global model. A privacy-preserving computation matching engine is deployed on and integrated into the trusted central node. It is used to encrypt and query profiles using attribute-based encryption technology and perform matching calculations in the ciphertext domain. The anonymous matching and authorization module is used to provide anonymous communication and a controllable information hierarchical decryption channel for both parties who are successfully matched.

2. The system according to claim 1, characterized in that, The personal dynamic skill profile is generated by fusing static attributes and dynamic behavioral data. The dynamic behavioral data includes at least one of authorized code repository contribution records, learning behavior data from online education platforms, or collaboration data from project collaboration tools. Skill proficiency and activity are quantified through a time-series model.

3. The system according to claim 1, characterized in that, The global model is updated through the following federated learning process: the trusted central node distributes the current global model to each local node; each local node calculates model parameter updates or gradients based on local data; each local node uploads the encrypted updates or gradients to the trusted central node. The trusted central node performs secure aggregation to generate a new version of the global model and then distributes it.

4. The system according to claim 3, characterized in that, The secure aggregation employs homomorphic encryption or secure multi-party computation techniques to ensure that the trusted central node cannot know the original update content of any local node during the aggregation process.

5. The system according to claim 1, characterized in that, The privacy-computing matching engine includes: The encryption unit is used to encrypt the image into a ciphertext index using an attribute-based encryption algorithm, and its decryption strategy is associated with the attributes of the image. The query processing unit is used to convert the queryer's search criteria into search tokens that are bound to their attributes; The ciphertext matching unit is used to calculate the matching degree between the search token and the ciphertext index without decryption.

6. The system according to claim 1 or 5, characterized in that, The anonymous matching and authorization module is configured to: in the initial matching stage, only return anonymous candidate identifiers and desensitized matching scores to the querying party; after both parties reach a preliminary agreement through the anonymous channel, the queried party authorizes the decryption of more detailed information layers in its profile level by level.

7. The system according to claim 1, characterized in that, It also includes a market trend analysis module, which generates macro skills supply and demand heatmaps, salary bandwidth and liquidity analysis reports based on anonymized statistical information uploaded by each local node and processed with differential privacy, and feeds them back to each local node.

8. A method for sharing human resource information based on big data, characterized in that, Applied to a system including a trusted central node and multiple local nodes, the method includes: S1. Each local node uses local data to participate in federated learning coordinated by the trusted central node, and jointly trains and maintains a global dynamic profile model. S2. Each local node generates a dynamic profile of an individual or organization based on the global model and its own data, and uploads it to the encrypted index library of the trusted central node after encrypting it using attribute-based encryption technology. S3. When a local node initiates a query, the trusted central node receives its encrypted query request, performs privacy-preserving matching calculations in the encrypted index, and returns an anonymous result list that conforms to the attribute policy. S4. Through the anonymous communication mechanism provided by the system, a connection is established between the querying party and the candidates in the result list, and more information is exchanged step by step based on the candidates' active authorization.

9. The method according to claim 8, characterized in that, In step S1, the federated learning process specifically includes: S5. The trusted central node initializes the global model parameters and broadcasts them to all participating nodes; S6. Each participating node trains the model on its local dataset and calculates the parameter differences or gradients. S7. Each participating node encrypts the parameter differences or gradients before uploading them; S8. The trusted central node performs secure aggregation of encrypted information and updates global model parameters. S9. Repeat S6-S8 until the model converges.

10. The method according to claim 8, characterized in that, In step S3, the privacy-preserving matching calculation specifically involves: converting the query request into a set of attribute policy-related search tokens, filtering out ciphertext indexes that meet the conditions through the ciphertext policy matching mechanism of the attribute-based encryption algorithm, and calculating the similarity of the portrait feature vectors in the encrypted vector space.

Images