Federal learning-based enterprise management consultation data privacy protection method and platform

Abstract

The invention discloses an enterprise management consultation data privacy protection method and platform based on federal learning, and relates to the field of management consultation and data privacy protection. According to the method, data localization isolation calculation is realized through a distributed calculation framework and a federated learning algorithm, the multi-party cooperation efficiency is improved on the premise of ensuring privacy security, and the data sharing breadth and depth are promoted; a contribution value is recorded by means of a block chain, and a dynamic income distribution rule is combined, so that fair and transparent income distribution is ensured, trust construction is enhanced through a distributed query interface, and enterprise participation motivation is stimulated; based on trust feedback, an incentive parameter is dynamically adjusted, an incentive mechanism is optimized to adapt to enterprise demand changes, an encryption protocol is continuously enhanced, two-way improvement of safety and efficiency is achieved, and technical support is provided for digital transformation in the field of enterprise management consultation.

Description

Technical Field

[0001] This invention relates to the fields of management consulting and data privacy protection technology, specifically to a data privacy protection method and platform for enterprise management consulting based on federated learning. Background Technology

[0002] In the field of enterprise management consulting, data sharing and analysis are crucial for driving industry innovation and improving decision-making efficiency. This area not only concerns the enhancement of core competitiveness but also directly impacts the overall digital transformation process of the industry. However, as data is a core asset of enterprises, the balance between privacy protection and shared utilization remains a critical issue. Currently, although many solutions attempt to protect privacy through traditional data encryption or centralized management, these methods often overlook the issues of trust building in data sharing and the uneven distribution of data value and contribution among different enterprises. This limitation leads to insufficient motivation for enterprises to participate in collaboration and makes it difficult to ensure transparency in data use and strict enforcement of rules, thus restricting the breadth and depth of data sharing.

[0003] Against this backdrop, the core challenges facing this field are gradually becoming apparent. First, achieving data privacy protection requires multi-party collaboration without exposing original information. This necessitates ensuring secure interaction between localized data processing and model parameters, and the complexity of this step often leads to a decline in collaboration efficiency. As the scale of collaboration expands, how to incentivize more companies to participate while ensuring security, and fairly distribute benefits based on their contributions, becomes an even more challenging issue. These two factors are closely related: the technical barriers to data security directly affect companies' willingness to participate, while the lack of effective incentives further weakens the sustainability of collaboration. Summary of the Invention

[0004] The purpose of this invention is to provide a data privacy protection method and platform for enterprise management consulting based on federated learning. It enables efficient multi-party collaboration while ensuring data privacy and security, and solves the problems of privacy leakage, insufficient incentives and low collaboration efficiency in traditional solutions through a fair and transparent incentive mechanism and a dynamically optimized trust feedback system.

[0005] The objective of this invention can be achieved through the following technical solutions: This application provides a data privacy protection method for enterprise management consulting based on federated learning, including the following steps; By using a pre-established distributed computing framework, a decentralized data processing environment is built to meet the local data processing needs of participating enterprises. Each original dataset is processed in isolation to obtain a preliminary set of encrypted features. By employing a joint learning algorithm to interact with the model parameters of the data features of each enterprise, multi-party collaborative training is completed without exposing the original information, the parameter update results of the shared model are determined, and a blockchain-based recording mechanism is constructed to bind the contribution value of each model training with the data value assessment, thereby obtaining the contribution weight record of each participant. Based on the preset profit distribution rules, the profit share of each participant is dynamically calculated to obtain the distribution result under the fair incentive mechanism, and a transparent rule execution log is generated. The profit distribution calculation process is linked with the contribution weight record to determine the basis for verifying the profit transparency of each participant. Build a distributed query interface that allows participants to access their respective revenue and contribution data and obtain real-time trust building feedback as the collaboration scales up. By combining the dynamic changes in enterprises' willingness to participate, the incentive parameters in the collaboration framework are adjusted. If the feedback data is lower than the preset trust threshold, the incentive mechanism is optimized to obtain an updated participation incentive plan. The weights of the benefit distribution for each participant in the incentive scheme are recalculated, and the adjusted parameters are applied to the next round of multi-party collaborative training to determine new criteria for improving collaborative efficiency. The impact of security technology barriers on collaboration is continuously monitored, and the model parameter interaction process is strengthened by adopting an encrypted interaction protocol to obtain the final secure collaboration optimization results.

[0006] Furthermore, a preliminary set of encryption features is obtained, specifically including: By pre-building a framework, the data processing environment is initialized to meet the needs of distributed computing and localized processing, the distribution of the original datasets of participating enterprises is obtained, and the data processing task allocation scheme for each node is determined. In a decentralized environment, isolated computation is performed on the original dataset. Data is divided using sharding technology, and localized processing is performed on each node. Encryption technology is used to protect the data fragments, generating encrypted data fragment groups. In conjunction with the feature extraction process, the preset feature extraction rules are used to perform feature mapping on each data segment group to obtain the corresponding preliminary feature set. The random forest algorithm is then used to predict and fill in the missing parts to obtain the complete feature set. Feature integration is performed within a distributed computing framework. Consistency verification is conducted on the integrated feature data to determine whether the feature set meets the preset standards. The feature set is then encrypted a second time through a data security mechanism to generate the final encrypted feature set.

[0007] Furthermore, the parameter update results of the shared model are determined, specifically including: A joint learning approach is used to process encryption features and data features, construct a preliminary feature matrix, determine the distribution characteristics of the feature matrix, and apply a joint learning algorithm to perform interactive calculation of model parameters on multi-party data features to obtain a preliminary set of model parameters. The training process of multi-party collaboration is iteratively optimized based on the initial model parameter set to obtain the optimized parameter update values. If the optimized parameter update values ​​do not match the preset threshold range, the data features in the training process are redistributed to determine whether the convergence condition is met. When the convergence condition is met, a shared model is constructed based on the updated parameter values, the structural framework of the shared model is obtained, the parameter update and output results are finally verified, the final parameter update results are generated, and the final model configuration of multi-party collaborative training is output.

[0008] Furthermore, obtain the contribution weight record of each participant, specifically including: The system retrieves model parameter update data uploaded by each participant, calculates a unique identifier for the parameter update data using a hash function, stores it in the blockchain distributed ledger to obtain parameter update records, aggregates the parameter update records using a preset federated learning algorithm, and calculates global model parameters. The gradient calculation method is used to evaluate the contribution of the parameters uploaded by each participant to the model performance, and the contribution value of each participant is obtained. When the contribution value is greater than the preset threshold, the quality weight of the data provided by the participants is calculated through the data value assessment model to obtain the data value assessment result. By binding contribution values ​​with data value assessment results through smart contracts, contribution weight records of each participant are generated. The integrity of the records is then verified through a consensus mechanism to obtain the final contribution weight of each participant. The training parameters of the shared model are updated using a weighted average algorithm to obtain the optimized shared model state.

[0009] Furthermore, the distribution results under a fair incentive mechanism are obtained, specifically including: Obtain the contribution weight data of each participant, extract quantitative indicators from the preset weight records, and map the contribution weight set using preset revenue distribution rules to determine the initial revenue share distribution ratio. If the initial profit share allocation ratio meets the fair incentive constraint, the allocation ratio is output; if not, the weight coefficients are adjusted by a linear regression algorithm, and the profit share of each participant is dynamically calculated by combining the total profit data to generate a profit share set. By comparing the deviation value between the profit share set and the preset rules, it is determined whether the allocation result conforms to the fair incentive mechanism, and the deviation assessment result is obtained. When the deviation assessment result is less than the preset threshold, the final allocation result is output. When the profit share is greater than or equal to the preset threshold, the gradient descent algorithm is used to iteratively optimize the profit share to obtain the final allocation result. Then, the profit share of each participant is extracted, a structured allocation record is generated, and stored in the allocation result database.

[0010] Furthermore, the criteria for verifying the transparency of each participant's earnings are determined, specifically including: The contribution data of each participant is obtained, the contribution weights are extracted from the preset contribution evaluation model, the weight value of each participant is determined, and the income distribution of each participant is calculated using a linear weighted algorithm to obtain the preliminary income results. If the initial profit results are consistent with the preset distribution rules, a profit distribution record is generated; if they are inconsistent, the weight values ​​are adjusted and recalculated to determine the final profit distribution. A transparent log containing the weight values, calculation process, and distribution results is generated to record the basis for each participant's profit. A hash algorithm is used to encrypt the transparent logs, generating a unique identifier to ensure the logs' immutability. By storing encrypted transparent logs using blockchain technology, obtaining the storage address, obtaining a verifiable log access path, generating a profit verification link for each participant, and determining the basis for the transparency of each participant's profits.

[0011] Furthermore, obtain real-time trust-building feedback, specifically including: By constructing a distributed query interface, revenue data and contribution data are obtained from the data storage nodes of the participants. A pre-established classification model is used to classify the revenue data and contribution data and determine the data grouping after classification. If there are outliers in the categorized data groups, they are filtered by a preset threshold to obtain filtered data groups. Real-time feedback information is then obtained to determine whether the feedback information meets the trust building standards. When the real-time feedback information meets the standard, the corresponding revenue data and contribution data are pushed to the participants through the distributed query interface, the push result is determined, the access records of the participants are obtained, and it is determined whether the access records are consistent with the trend of changes in the scale of collaboration. When the trend of access records and collaboration scale changes are consistent, the trust building status is updated through data access logs to obtain the final trust feedback information.

[0012] Furthermore, the updated participation incentive scheme includes: By obtaining enterprise participation intention data through trust feedback, and using time series analysis, the trend of participation intention changes is determined. When the trend of participation intention changes is lower than a preset threshold, a trust score is calculated based on the trust feedback data to obtain the trust score distribution. A linear regression model is used to predict the adjustment range of incentive parameters, determine the parameter adjustment scheme, update the collaboration framework configuration through the parameter adjustment scheme, obtain the updated incentive mechanism, and judge the stability of the incentive mechanism. When the stability of the incentive mechanism is higher than the preset standard, a new participation incentive plan is generated, feedback data on the enterprise's willingness to participate is obtained, and the relationship between the feedback data and the trust threshold is determined. When the feedback data is lower than the trust threshold, the incentive parameters are dynamically adjusted through feedback to obtain an optimized incentive plan.

[0013] Furthermore, new criteria for improving collaboration efficiency will be identified, including: By acquiring the latest incentive scheme data and processing parameter adjustments, the preliminary profit distribution results of each participant are obtained. Using a preset distribution weight model, the profit weight value of each participant is calculated, and the updated weight distribution is determined. When the weight distribution deviates from the result of the previous training round by more than a preset threshold, the benefit calculation is calibrated through the information processing stage. Combined with the training data from multi-party collaboration, a logistic regression model is used to predict the efficiency improvement of collaborative training. Determine whether the prediction results meet the collaboration criteria, obtain new training efficiency data, compare the new training efficiency data with historical data, determine the final basis for improving collaboration efficiency, optimize the parameter adjustment direction for the next round of multi-party collaboration, and obtain the optimized collaboration scheme data.

[0014] This application provides a data privacy protection platform for enterprise management consulting based on federated learning, used to implement data privacy protection methods for enterprise management consulting based on federated learning, including: The distributed computing framework module is responsible for building a decentralized data processing environment, performing localized isolated computation on the original datasets of participating enterprises, determining data distribution and task allocation through initialization configuration, segmenting and encrypting data using sharding technology, filling missing features with feature extraction and random forest algorithms, and finally completing feature integration and secondary encryption to generate an encrypted feature set. The joint learning and blockchain recording module enables multi-party model parameter interaction through joint learning algorithms, completing collaborative training without exposing the original data. At the same time, it uses blockchain to record the model parameter update data of each participant, and binds the contribution value and data value assessment to the smart contract through a hash function to generate an immutable contribution weight record. The dynamic revenue distribution module dynamically calculates the revenue share of each participant based on preset rules; it optimizes the distribution ratio through linear regression and gradient descent algorithms to ensure fairness, and generates a transparent rule execution log, linking the calculation process with contribution records to provide a basis for verifying revenue transparency. The distributed query and trust feedback module builds a distributed query interface, allowing participants to access revenue and contribution data; it processes data and filters outliers through a classification model, judges in real time whether the feedback information meets the trust standards, updates the trust status based on the access records, and provides dynamic feedback for building collaborative trust. The incentive mechanism optimization module analyzes the changing trends of corporate participation intentions and adjusts incentive parameters through time series and linear regression models; when the trust score is below the threshold, the incentive mechanism is optimized and an updated participation incentive plan is generated. The secure collaboration enhancement module continuously monitors the impact of security technology barriers on collaboration, strengthens the model parameter interaction process through encrypted interaction protocols, classifies and handles technical barriers and protects against security risks, optimizes encryption strength by combining logistic regression algorithms until the security threshold is met, and finally integrates efficiency improvement results to output secure collaboration optimization results.

[0015] The beneficial effects of this invention are as follows: By leveraging a distributed computing framework and federated learning algorithms, localized data processing and isolated computation are achieved, ensuring that data can complete multi-party collaborative training without leaving the enterprise's local area. This effectively addresses the risk of privacy leaks in traditional data sharing. At the same time, by optimizing the collaboration process, the efficiency of multi-party collaboration is significantly improved. Enterprises do not need to worry about the leakage of core data assets and can participate more actively in data sharing, thereby promoting the breadth and depth of data sharing and providing solid technical support for the digital transformation of industries. By using blockchain technology to record the contribution values ​​and data value assessments of model training, an immutable contribution weight record is generated. Combined with dynamic revenue distribution rules, the fairness and transparency of revenue distribution are ensured. At the same time, a distributed query interface is built, allowing participants to access revenue and contribution data in real time, further enhancing trust building. This mechanism not only solves the problem of insufficient corporate participation due to a lack of effective incentives in traditional methods, but also enhances corporate trust and strengthens the participants' motivation to collaborate through transparent log recording and real-time feedback. Meanwhile, the immutability of blockchain ensures the transparency and credibility of the distribution process. By combining a distributed query interface and a trust feedback mechanism with dynamic adjustment of incentive parameters, this dynamic adjustment mechanism solves the problem of fixed incentive mechanisms in traditional methods, which are difficult to adapt to the dynamic needs of enterprises. By continuously optimizing incentive parameters, the stability and effectiveness of the incentive mechanism are ensured, further enhancing enterprises' willingness to participate and their sense of trust. Ultimately, this optimization mechanism not only improves the enthusiasm of enterprises to participate, but also improves collaboration efficiency while ensuring data security by continuously monitoring security technology barriers and strengthening encrypted interaction protocols, achieving a two-way optimization of security and efficiency. Attached Figure Description

[0016] To better understand and implement this application, the technical solution is described in detail below with reference to the accompanying drawings.

[0017] Figure 1 A flowchart illustrating the data privacy protection method for enterprise management consulting based on federated learning provided in Embodiment 1 of this application; Figure 2 A schematic diagram illustrating the process of obtaining the contribution weight record of each participant using the federated learning-based enterprise management consulting data privacy protection method provided in Embodiment 1 of this application; Figure 3 A flowchart illustrating the process of obtaining the allocation results under a fair incentive mechanism using the data privacy protection method for enterprise management consulting based on federated learning provided in Embodiment 1 of this application; Figure 4 This is a schematic diagram of the structure of the enterprise management consulting data privacy protection platform based on federated learning provided in Embodiment 2 of this application. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, exemplary embodiments will be described in detail below, examples of which are illustrated in the accompanying drawings. In the following description relating to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of methods and systems consistent with some aspects of this application as detailed in the appended claims.

[0019] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0020] The following detailed description of the specific implementation methods, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided in detail.

[0021] Example 1

[0022] Please see Figures 1-3 This embodiment provides a data privacy protection method for enterprise management consulting based on federated learning, including the following steps; S1. Through a pre-established distributed computing framework, a decentralized data processing environment is built to meet the local data processing needs of participating enterprises. In this environment, each original dataset is processed in isolation to obtain a preliminary set of encrypted features.

[0023] Furthermore, a preliminary set of encryption features is obtained, specifically including: By pre-building a framework, the data processing environment is initialized to meet the needs of distributed computing and localized processing, the distribution of the original datasets of participating enterprises is obtained, and the data processing task allocation scheme for each node is determined. According to the allocation scheme, the original dataset is subjected to isolated computation in a decentralized environment. The data is divided using sharding technology to obtain a set of initially divided data fragments. Localized processing is performed on each node, and encryption technology is used to protect the data fragments to generate an encrypted data fragment group. In conjunction with the feature extraction process, using preset feature extraction rules, feature mapping is performed on each data segment group to obtain the corresponding preliminary feature set. When the completeness of the preliminary feature set does not reach the preset threshold, the feature set is supplemented and calculated. The random forest algorithm is used to predict and fill in the missing parts to obtain the complete feature set. Feature integration is performed within the distributed computing framework. Consistency verification is then performed on the integrated feature data to determine whether the feature set meets the preset standards. If the feature set meets the preset standards, the feature set is then encrypted a second time through a data security guarantee mechanism to generate the final encrypted feature set.

[0024] Specifically, by constructing a decentralized distributed computing framework and employing methods such as sharding, encryption, feature extraction, and supplementary computation, localized data processing and secure sharing have been achieved. This effectively protects enterprise data privacy while improving data integrity and availability, significantly enhancing the efficiency and security of multi-party collaboration, providing enterprises with more reliable technical support in the data sharing process, and promoting the digital transformation and innovative development of the enterprise management consulting field.

[0025] S2. Based on the initial set of encrypted features, a joint learning algorithm is used to interact with the model parameters of the data features of each enterprise. Multi-party collaborative training is completed without exposing the original information. The parameter update results of the shared model are determined. Then, a blockchain-based recording mechanism is constructed to bind the contribution value of each model training with the data value assessment and obtain the contribution weight record of each participant.

[0026] Furthermore, the parameter update results of the shared model are determined, specifically including: A joint learning approach is used to process encryption features and data features. An initial set of encryption features is obtained from various enterprises, a preliminary feature matrix is ​​constructed, the distribution characteristics of the feature matrix are determined, and a joint learning algorithm is applied to perform interactive calculation of model parameters on the data features from multiple parties to obtain a preliminary set of model parameters. The training process of multi-party collaboration is iteratively optimized based on the initial model parameter set to obtain the optimized parameter update values. If the optimized parameter update values ​​do not match the preset threshold range, the data features in the training process are redistributed to determine whether the convergence condition is met. When the convergence condition is met, a shared model is constructed based on the updated parameter values ​​to obtain the structural framework of the shared model. The parameter update and output results are then verified to determine the applicability of the shared model. Finally, the parameter update results are generated, and the final model configuration for multi-party collaborative training is output.

[0027] Specifically, by employing federated learning algorithms and blockchain technology, multi-party collaborative training was achieved without exposing the original data, effectively resolving the conflict between data privacy protection and model sharing. The federated learning algorithm ensures secure interaction of data features from various enterprises and efficient updates of model parameters, improving the model's accuracy and generalization ability. Simultaneously, the blockchain-based recording mechanism binds the contribution value of model training to data value assessment, generating immutable contribution weight records. This provides a reliable basis for fair incentive mechanisms, enhancing trust among participating enterprises and improving the efficiency and transparency of multi-party collaboration, thus promoting data sharing and collaborative innovation in the field of enterprise management consulting.

[0028] Furthermore, obtain the contribution weight record of each participant, specifically including: S21. Obtain the model parameter update data uploaded by each participant, calculate the unique identifier of the parameter update data through a hash function, store it in the blockchain distributed ledger to obtain the parameter update record, aggregate the parameter update record through a preset federated learning algorithm, calculate the global model parameters, and obtain the latest state of the shared model. S22. The gradient calculation method is used to evaluate the contribution of the parameters uploaded by each participant to the model performance, and the contribution value of each participant is obtained. When the contribution value is greater than the preset threshold, the quality weight of the data provided by the participants is calculated through the data value assessment model to obtain the data value assessment result. S23. By binding the contribution value with the data value assessment result through smart contracts, a contribution weight record of each participant is generated and stored in the blockchain to obtain an immutable weight record. The integrity of the record is then verified through a consensus mechanism to obtain the final contribution weight of each participant. The training parameters of the shared model are updated using a weighted average algorithm to obtain the optimized shared model state.

[0029] Specifically, it achieves precise quantification and transparent recording of the contributions of each participant. Utilizing hash functions and blockchain technology, it ensures the immutability and traceability of model parameter update records, enhancing the trust foundation of the data sharing process. Simultaneously, through gradient calculation and data value assessment models, it scientifically measures the contributions and data quality of each participant, providing a quantitative basis for fair incentives. Finally, through smart contracts and consensus mechanisms, it automatically generates and verifies the contribution weight records of each participant, ensuring the transparency and fairness of the incentive mechanism and further enhancing the enthusiasm and sustainability of multi-party collaboration.

[0030] S3. By recording contribution weights and combining them with preset revenue distribution rules, the revenue share of each participant is dynamically calculated to obtain the distribution result under the fair incentive mechanism. A transparent rule execution log is generated, and the revenue distribution calculation process is linked with the contribution weight records to determine the basis for verifying the revenue transparency of each participant.

[0031] Furthermore, the distribution results under a fair incentive mechanism are obtained, specifically including: S31. Obtain the contribution weight data of each participant, extract quantitative indicators from the preset weight records to obtain the contribution weight set, and use the preset revenue distribution rules to map the contribution weight set to determine the initial revenue share distribution ratio. S32. If the initial profit share allocation ratio meets the fair incentive constraint, the allocation ratio is output; if not, the weight coefficients are adjusted by linear regression algorithm to obtain the optimized allocation ratio. Combined with the total profit data, the profit share of each participant is dynamically calculated to generate a profit share set. S33. By comparing the deviation value between the profit share set and the preset rules, determine whether the allocation result conforms to the fair incentive mechanism and obtain the deviation assessment result. When the deviation assessment result is less than the preset threshold, output the final allocation result. S34. When the profit share is greater than or equal to the preset threshold, the gradient descent algorithm is used to iteratively optimize the profit share to obtain the final allocation result. Then, the profit share of each participant is extracted, a structured allocation record is generated, and stored in the allocation result database.

[0032] Specifically, a fair revenue distribution mechanism based on contribution weights was implemented, effectively addressing the issues of opaque and unfair revenue distribution in multi-party collaborations. By utilizing pre-defined revenue distribution rules and dynamic adjustment algorithms (such as linear regression and gradient descent), the scientific rigor and adaptability of the revenue distribution ratio were ensured. Simultaneously, transparent rule execution logs and structured distribution records were generated, providing each participant with clear verification of revenue transparency. This mechanism not only enhances enterprises' enthusiasm for participating in data sharing but also improves the trust and sustainability of the entire collaborative ecosystem, providing a solid economic incentive foundation for multi-party collaboration.

[0033] Furthermore, the criteria for verifying the transparency of each participant's earnings are determined, specifically including: The contribution data of each participant is obtained, the contribution weights are extracted from the preset contribution evaluation model, the weight value of each participant is determined, and the income distribution of each participant is calculated using a linear weighted algorithm to obtain the preliminary income results. If the initial profit results are consistent with the preset distribution rules, a profit distribution record is generated; if they are inconsistent, the weight values ​​are adjusted and recalculated to determine the final profit distribution. A transparent log containing the weight values, calculation process, and distribution results is generated to record the basis for each participant's profit. A hash algorithm is used to encrypt the transparent logs, generating a unique identifier to ensure the logs' immutability. By storing encrypted transparent logs using blockchain technology, obtaining the storage address, obtaining a verifiable log access path, generating a profit verification link for each participant, and determining the basis for the transparency of each participant's profits.

[0034] Specifically, it further enhances the transparency and immutability of revenue distribution. Utilizing a linear weighted algorithm and hash encryption technology, it ensures that the revenue distribution process for each participant is clear and traceable. Simultaneously, it stores encrypted transparent logs through blockchain technology, generating verifiable revenue verification links. This mechanism provides participants with a solid basis for revenue transparency, greatly enhancing trust among all parties in the collaboration, further consolidating the trust foundation for multi-party cooperation, and promoting the healthy development of the data sharing ecosystem.

[0035] S4. To verify the transparency of revenue, a distributed query interface is built, allowing participants to access their respective revenue and contribution data and obtain real-time trust building feedback as the scale of collaboration expands.

[0036] Furthermore, obtain real-time trust-building feedback, specifically including: By constructing a distributed query interface, revenue data and contribution data are obtained from the data storage nodes of the participants. A pre-established classification model is used to classify the revenue data and contribution data and determine the data grouping after classification. If there are outliers in the categorized data groups, they are filtered by a preset threshold to obtain filtered data groups. Real-time feedback information is then obtained to determine whether the feedback information meets the trust building standards. When the real-time feedback information meets the standard, the corresponding revenue data and contribution data are pushed to the participants through the distributed query interface, the push result is determined, the access records of the participants are obtained, and it is determined whether the access records are consistent with the trend of changes in the scale of collaboration. When the trend of access records and collaboration scale changes are consistent, the trust building status is updated through data access logs to obtain the final trust feedback information.

[0037] Specifically, by constructing a distributed query interface and combining it with a real-time feedback mechanism, this system enables each participant to quickly and conveniently access their own revenue and contribution data as the collaboration scales up. This mechanism not only improves data accessibility and real-time performance but also ensures data accuracy and reliability through outlier filtering and trust standard verification. Ultimately, real-time push notifications and trust status updates enhance the participants' sense of trust, promote the stability and sustainability of the collaborative ecosystem, and provide a solid foundation of trust for multi-party collaboration.

[0038] S5. By building feedback through trust and combining it with the dynamic changes in the enterprise's willingness to participate, the incentive parameters in the collaboration framework are adjusted. If the feedback data is lower than the preset trust threshold, the incentive mechanism is optimized to obtain an updated participation incentive plan.

[0039] Furthermore, the updated participation incentive scheme includes: By obtaining enterprise participation intention data through trust feedback, and using time series analysis, the trend of participation intention changes is determined. When the trend of participation intention changes is lower than a preset threshold, a trust score is calculated based on the trust feedback data to obtain the trust score distribution. A linear regression model is used to predict the adjustment range of incentive parameters, determine the parameter adjustment scheme, update the collaboration framework configuration through the parameter adjustment scheme, obtain the updated incentive mechanism, and judge the stability of the incentive mechanism. When the stability of the incentive mechanism is higher than the preset standard, a new participation incentive plan is generated, the incentive plan output is obtained, feedback data on the enterprise's willingness to participate is obtained, and the relationship between the feedback data and the trust threshold is determined. When the feedback data is lower than the trust threshold, the incentive parameters are dynamically adjusted through feedback to obtain an optimized incentive plan.

[0040] Specifically, by dynamically adjusting incentive parameters and combining them with trust-building feedback, the incentive mechanism can be optimized in real time based on changes in corporate participation willingness. This mechanism not only improves the adaptability and effectiveness of the incentive scheme, but also ensures its stability and sustainability through dynamic adjustments to trust scores and feedback data. Ultimately, this flexible incentive scheme can effectively enhance corporate participation and collaboration enthusiasm, further consolidate the trust foundation of the collaborative ecosystem, and promote the long-term stable development of multi-party collaboration.

[0041] S6. Based on the updated participation incentive scheme, recalculate the revenue distribution weights of each participant, apply the adjusted parameters to the next round of multi-party collaborative training, determine the new basis for improving collaborative efficiency, continuously monitor the impact of security technology barriers on collaboration, and strengthen the model parameter interaction process using an encrypted interaction protocol to obtain the final secure collaboration optimization results.

[0042] Furthermore, new criteria for improving collaboration efficiency will be identified, including: By acquiring the latest incentive scheme data and processing parameter adjustments, the preliminary profit distribution results of each participant are obtained. Using a preset distribution weight model, the profit weight value of each participant is calculated, and the updated weight distribution is determined. When the weight distribution deviates from the result of the previous training round by more than a preset threshold, the benefit calculation is calibrated through the information processing stage to obtain the calibrated weight data. Combined with the training data from multi-party collaboration, a logistic regression model is used to predict the efficiency improvement of collaborative training. If the prediction results meet the criteria for collaboration, the updated weights are applied to the next round of collaborative training to obtain new training efficiency data. The new training efficiency data is then compared with historical data to determine the final basis for improving collaborative efficiency. The direction of parameter adjustment for the next round of multi-party collaboration is then optimized to obtain the optimized collaboration scheme data.

[0043] Specifically, by dynamically adjusting the participation incentive scheme and optimizing the profit distribution weights, the profit distribution of each participant can be recalculated based on the latest incentive scheme and applied to the next round of multi-party collaborative training. Simultaneously, by continuously monitoring security technical barriers and employing encrypted interaction protocols to strengthen the model parameter interaction process, the security and stability of the collaboration are ensured. This mechanism not only improves collaboration efficiency but also further consolidates the sustainability of multi-party collaboration by predicting and optimizing training efficiency, providing efficient, secure, and reliable technical support for data sharing and collaborative innovation.

[0044] Furthermore, the final optimized results for secure collaboration are obtained, specifically including: By deploying monitoring tools to collect real-time data on security technical barriers in the collaboration process, we can obtain data on the impact of technical barriers on collaboration efficiency, determine the scope of impact and key nodes, and classify the specific manifestations of technical barriers based on the collected impact data using a pre-established analysis framework to obtain a set of barrier features after classification. Acquire data on interaction links related to the collaboration process, determine whether there are any security risks in the interaction links, and when a risk is detected, protect the risky links through encrypted interaction protocols and determine the status of the protected interaction links. By continuously monitoring the status of the protected interaction process, the changing trend of model parameters during the interaction process is obtained, and the security assessment results of parameter interaction are obtained. Then, the encryption strength of the interaction protocol is optimized and adjusted using a logistic regression algorithm to determine the adjusted encryption strategy configuration. By applying the adjusted encryption strategy configuration, the parameter interaction links in the collaboration process are strengthened in real time, the enhanced secure interaction data is obtained, and it is determined whether the preset security threshold is reached. If not, the encryption strategy is adjusted cyclically until the threshold requirement is met. Then, the optimization results of the improved collaboration efficiency are integrated to obtain the final secure collaboration output data.

[0045] Specifically, by monitoring and optimizing security barriers in the collaboration process in real time, potential security risks can be accurately identified and addressed, ensuring the security and stability of parameter interaction. Combined with encrypted interaction protocols and dynamically adjusted encryption strategies, the security of the collaboration process is further enhanced. At the same time, by continuously monitoring and optimizing collaboration efficiency, both security and efficiency are improved. Ultimately, this comprehensive security optimization mechanism provides a solid guarantee for multi-party collaboration, ensuring the efficient, stable, and sustainable development of the data sharing ecosystem.

[0046] Example 2

[0047] Please see Figure 4 This embodiment provides a data privacy protection platform for enterprise management consulting based on federated learning, used to implement a data privacy protection method for enterprise management consulting based on federated learning, including: The distributed computing framework module is responsible for building a decentralized data processing environment, performing localized isolated computation on the original datasets of participating enterprises, determining data distribution and task allocation through initialization configuration, segmenting and encrypting data using sharding technology, filling missing features with feature extraction and random forest algorithms, and finally completing feature integration and secondary encryption to generate an encrypted feature set. The joint learning and blockchain recording module, based on a set of encrypted features, enables multi-party model parameter interaction through a joint learning algorithm, completing collaborative training without exposing the original data. At the same time, it uses blockchain to record the model parameter update data of each participant, and binds the contribution value and data value assessment to the smart contract through a hash function to generate an immutable contribution weight record. The dynamic revenue allocation module dynamically calculates the revenue share of each participant based on contribution weight records and preset rules; it optimizes the allocation ratio through linear regression and gradient descent algorithms to ensure fairness, and generates a transparent rule execution log, linking the calculation process with contribution records to provide a basis for revenue transparency verification. The distributed query and trust feedback module builds a distributed query interface, allowing participants to access revenue and contribution data; it processes data and filters outliers through a classification model, judges in real time whether the feedback information meets the trust standards, updates the trust status based on the access records, and provides dynamic feedback for building collaborative trust. The incentive mechanism optimization module analyzes the changing trends of corporate participation intentions based on trust feedback data and adjusts incentive parameters through time series and linear regression models. When the trust score is below the threshold, the incentive mechanism is optimized to generate an updated participation incentive plan, balancing the enthusiasm of all parties for cooperation. The secure collaboration enhancement module continuously monitors the impact of security technology barriers on collaboration, strengthens the model parameter interaction process through encrypted interaction protocols, classifies and handles technical barriers and protects against security risks, optimizes encryption strength by combining logistic regression algorithms until the security threshold is met, and finally integrates efficiency improvement results to output secure collaboration optimization results.

[0048] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A data privacy protection method for enterprise management consulting based on federated learning, characterized by: Includes the following steps; By using a pre-established distributed computing framework, a decentralized data processing environment is built to meet the local data processing needs of participating enterprises. Each original dataset is processed in isolation to obtain a preliminary set of encrypted features. By employing a joint learning algorithm to interact with the model parameters of the data features of each enterprise, multi-party collaborative training is completed without exposing the original information, the parameter update results of the shared model are determined, and a blockchain-based recording mechanism is constructed to bind the contribution value of each model training with the data value assessment, thereby obtaining the contribution weight record of each participant. Based on the preset profit distribution rules, the profit share of each participant is dynamically calculated to obtain the distribution result under the fair incentive mechanism, and a transparent rule execution log is generated. The profit distribution calculation process is linked with the contribution weight record to determine the basis for verifying the profit transparency of each participant. Build a distributed query interface that allows participants to access their respective revenue and contribution data and obtain real-time trust building feedback as the collaboration scales up. By combining the dynamic changes in enterprises' willingness to participate, the incentive parameters in the collaboration framework are adjusted. If the feedback data is lower than the preset trust threshold, the incentive mechanism is optimized to obtain an updated participation incentive plan. The weights of the benefit distribution for each participant in the incentive scheme are recalculated, and the adjusted parameters are applied to the next round of multi-party collaborative training to determine new criteria for improving collaborative efficiency. The impact of security technology barriers on collaboration is continuously monitored, and the model parameter interaction process is strengthened by adopting an encrypted interaction protocol to obtain the final secure collaboration optimization results.

2. The data privacy protection method for enterprise management consulting based on federated learning according to claim 1, characterized in that: A preliminary set of encryption features is obtained, specifically including: By pre-building a framework, the data processing environment is initialized to meet the needs of distributed computing and localized processing, the distribution of the original datasets of participating enterprises is obtained, and the data processing task allocation scheme for each node is determined. In a decentralized environment, isolated computation is performed on the original dataset. Data is divided using sharding technology, and localized processing is performed on each node. Encryption technology is used to protect the data fragments, generating encrypted data fragment groups. In conjunction with the feature extraction process, the preset feature extraction rules are used to perform feature mapping on each data segment group to obtain the corresponding preliminary feature set. The random forest algorithm is then used to predict and fill in the missing parts to obtain the complete feature set. Feature integration is performed within a distributed computing framework. Consistency verification is conducted on the integrated feature data to determine whether the feature set meets the preset standards. The feature set is then encrypted a second time through a data security mechanism to generate the final encrypted feature set.

3. The data privacy protection method for enterprise management consulting based on federated learning according to claim 1, characterized in that: Determine the parameter update results of the shared model, specifically including: A joint learning approach is used to process encryption features and data features, construct a preliminary feature matrix, determine the distribution characteristics of the feature matrix, and apply a joint learning algorithm to perform interactive calculation of model parameters on multi-party data features to obtain a preliminary set of model parameters. The training process of multi-party collaboration is iteratively optimized based on the initial model parameter set to obtain the optimized parameter update values. If the optimized parameter update values ​​do not match the preset threshold range, the data features in the training process are redistributed to determine whether the convergence condition is met. When the convergence condition is met, a shared model is constructed based on the updated parameter values, the structural framework of the shared model is obtained, the parameter update and output results are finally verified, the final parameter update results are generated, and the final model configuration of multi-party collaborative training is output.

4. The data privacy protection method for enterprise management consulting based on federated learning according to claim 1, characterized in that: Obtain the contribution weight record for each participant, specifically including: The system retrieves model parameter update data uploaded by each participant, calculates a unique identifier for the parameter update data using a hash function, stores it in the blockchain distributed ledger to obtain parameter update records, aggregates the parameter update records using a preset federated learning algorithm, and calculates global model parameters. The gradient calculation method is used to evaluate the contribution of the parameters uploaded by each participant to the model performance, and the contribution value of each participant is obtained. When the contribution value is greater than the preset threshold, the quality weight of the data provided by the participants is calculated through the data value assessment model to obtain the data value assessment result. By binding contribution values ​​with data value assessment results through smart contracts, contribution weight records of each participant are generated. The integrity of the records is then verified through a consensus mechanism to obtain the final contribution weight of each participant. The training parameters of the shared model are updated using a weighted average algorithm to obtain the optimized shared model state.

5. The data privacy protection method for enterprise management consulting based on federated learning according to claim 1, characterized in that: The distribution results obtained under a fair incentive mechanism specifically include: Obtain the contribution weight data of each participant, extract quantitative indicators from the preset weight records, and map the contribution weight set using preset revenue distribution rules to determine the initial revenue share distribution ratio. If the initial profit share allocation ratio meets the fair incentive constraint, the allocation ratio is output; if not, the weight coefficients are adjusted by a linear regression algorithm, and the profit share of each participant is dynamically calculated by combining the total profit data to generate a profit share set. By comparing the deviation value between the profit share set and the preset rules, it is determined whether the allocation result conforms to the fair incentive mechanism, and the deviation assessment result is obtained. When the deviation assessment result is less than the preset threshold, the final allocation result is output. When the profit share is greater than or equal to the preset threshold, the gradient descent algorithm is used to iteratively optimize the profit share to obtain the final allocation result. Then, the profit share of each participant is extracted, a structured allocation record is generated, and stored in the allocation result database.

6. The data privacy protection method for enterprise management consulting based on federated learning according to claim 1, characterized in that: The criteria for verifying the transparency of each participant's earnings should be determined, specifically including: The contribution data of each participant is obtained, the contribution weights are extracted from the preset contribution evaluation model, the weight value of each participant is determined, and the income distribution of each participant is calculated using a linear weighted algorithm to obtain the preliminary income results. If the initial profit results are consistent with the preset distribution rules, a profit distribution record is generated; if they are inconsistent, the weight values ​​are adjusted and recalculated to determine the final profit distribution. A transparent log containing the weight values, calculation process, and distribution results is generated to record the basis for each participant's profit. A hash algorithm is used to encrypt the transparent logs, generating a unique identifier to ensure the logs' immutability. By storing encrypted transparent logs using blockchain technology, obtaining the storage address, obtaining a verifiable log access path, generating a profit verification link for each participant, and determining the basis for the transparency of each participant's profits.

7. The data privacy protection method for enterprise management consulting based on federated learning according to claim 1, characterized in that: Obtain real-time trust-building feedback, specifically including: By constructing a distributed query interface, revenue data and contribution data are obtained from the data storage nodes of the participants. A pre-established classification model is used to classify the revenue data and contribution data and determine the data grouping after classification. If there are outliers in the categorized data groups, they are filtered by a preset threshold to obtain filtered data groups. Real-time feedback information is then obtained to determine whether the feedback information meets the trust building standards. When the real-time feedback information meets the standard, the corresponding revenue data and contribution data are pushed to the participants through the distributed query interface, the push result is determined, the access records of the participants are obtained, and it is determined whether the access records are consistent with the trend of changes in the scale of collaboration. When the trend of access records and collaboration scale changes are consistent, the trust building status is updated through data access logs to obtain the final trust feedback information.

8. The data privacy protection method for enterprise management consulting based on federated learning according to claim 1, characterized in that: The updated participation incentive plan includes: By obtaining enterprise participation intention data through trust feedback, and using time series analysis, the trend of participation intention changes is determined. When the trend of participation intention changes is lower than a preset threshold, a trust score is calculated based on the trust feedback data to obtain the trust score distribution. A linear regression model is used to predict the adjustment range of incentive parameters, determine the parameter adjustment scheme, update the collaboration framework configuration through the parameter adjustment scheme, obtain the updated incentive mechanism, and judge the stability of the incentive mechanism. When the stability of the incentive mechanism is higher than the preset standard, a new participation incentive plan is generated, feedback data on the enterprise's willingness to participate is obtained, and the relationship between the feedback data and the trust threshold is determined. When the feedback data is lower than the trust threshold, the incentive parameters are dynamically adjusted through feedback to obtain an optimized incentive plan.

9. The data privacy protection method for enterprise management consulting based on federated learning according to claim 1, characterized in that: The criteria for determining new ways to improve collaboration efficiency include: By acquiring the latest incentive scheme data and processing parameter adjustments, the preliminary profit distribution results of each participant are obtained. Using a preset distribution weight model, the profit weight value of each participant is calculated, and the updated weight distribution is determined. When the weight distribution deviates from the result of the previous training round by more than a preset threshold, the benefit calculation is calibrated through the information processing stage. Combined with the training data from multi-party collaboration, a logistic regression model is used to predict the efficiency improvement of collaborative training. Determine whether the prediction results meet the collaboration criteria, obtain new training efficiency data, compare the new training efficiency data with historical data, determine the final basis for improving collaboration efficiency, optimize the parameter adjustment direction for the next round of multi-party collaboration, and obtain the optimized collaboration scheme data.

10. A data privacy protection platform for enterprise management consulting based on federated learning, used to implement the data privacy protection method for enterprise management consulting based on federated learning as described in any one of claims 1-9, characterized in that: include: The distributed computing framework module is responsible for building a decentralized data processing environment, performing localized isolated computation on the original datasets of participating enterprises, determining data distribution and task allocation through initialization configuration, segmenting and encrypting data using sharding technology, filling missing features with feature extraction and random forest algorithms, and finally completing feature integration and secondary encryption to generate an encrypted feature set. The joint learning and blockchain recording module enables multi-party model parameter interaction through joint learning algorithms, completing collaborative training without exposing the original data. At the same time, it uses blockchain to record the model parameter update data of each participant, and binds the contribution value and data value assessment to the smart contract through a hash function to generate an immutable contribution weight record. The dynamic profit distribution module dynamically calculates the profit share of each participant based on preset rules. The allocation ratio is optimized by using linear regression and gradient descent algorithms to ensure fairness, and a transparent rule execution log is generated to link the calculation process with contribution records, providing a basis for verifying the transparency of benefits. The distributed query and trust feedback module builds a distributed query interface, allowing participants to access revenue and contribution data; it processes data and filters outliers through a classification model, judges in real time whether the feedback information meets the trust standards, updates the trust status based on the access records, and provides dynamic feedback for building collaborative trust. The incentive mechanism optimization module analyzes the changing trends of corporate participation intentions and adjusts incentive parameters through time series and linear regression models; when the trust score is below the threshold, the incentive mechanism is optimized and an updated participation incentive plan is generated. The secure collaboration enhancement module continuously monitors the impact of security technology barriers on collaboration, strengthens the model parameter interaction process through encrypted interaction protocols, classifies and handles technical barriers and protects against security risks, optimizes encryption strength by combining logistic regression algorithms until the security threshold is met, and finally integrates efficiency improvement results to output secure collaboration optimization results.

Images