A method, system, and storage medium for sharing cloud-based terminal data with a platform based on differential privacy.
By collecting, preprocessing, and filtering user terminal data in a cloud-based terminal data and platform sharing system, and configuring a differential privacy optimizer to train a model, the contradiction between data privacy protection and sharing efficiency is resolved, achieving both privacy protection and the security and trustworthiness of data sharing.
Patent Information
- Application Number
- CN202510840133.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-23
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2045-06-23
AI Technical Summary
Existing technologies struggle to protect data privacy while allowing fine-grained analysis and processing, and to ensure both data sharing efficiency and privacy security in multi-party data sharing scenarios.
By collecting operational data from user terminals, preprocessing it, and sending it to the cloud platform, usable data is filtered and stored. A differential privacy optimizer is configured to train a differential privacy model, which is then processed into privacy data. Parameters are adjusted during data transmission to resist attacks.
It enables privacy protection in large-scale dataset sharing scenarios, ensuring that individual data is not identified, improving the credibility and security of data sharing, and is suitable for sharing data between cloud terminals and platforms.
Smart Images

Figure CN120675772B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of cloud data processing, and in particular to a method, system, and storage medium for sharing cloud terminal data with a platform based on differential privacy. Background Technology
[0002] With the development of cloud computing technology, more and more organizations and individuals are migrating their data storage and processing tasks to cloud environments. While cloud environments offer numerous advantages such as elastic computing and on-demand scaling, they also bring challenges related to data privacy and security. Because data storage and processing occur on the infrastructure of cloud service providers, data privacy protection becomes complex. Existing technologies typically employ encryption algorithms such as AES and RSA to ensure data security during transmission and storage. They also guarantee data privacy and prevent unauthorized tampering through authentication mechanisms, data access controls, and data integrity protection. However, existing technologies struggle to simultaneously protect data privacy and allow for fine-grained data analysis. Although de-identification techniques can hide personally identifiable information, it can still be re-identified in complex attack scenarios. Furthermore, in multi-party data sharing scenarios, it is difficult to simultaneously guarantee data sharing efficiency and privacy security. Summary of the Invention
[0003] The purpose of this invention is to provide a method and system for sharing cloud terminal data with a platform based on differential privacy. This method collects and preprocesses operational data from user terminals to obtain preprocessed operational data, which is then sent to the cloud platform based on the operational status of the user terminals. This ensures comprehensive collection of user terminal data and a richness of data sources for the cloud platform. The preprocessed operational data received from the cloud platform is filtered for usable operational data and stored in the cloud platform's storage space. Furthermore, based on data acquisition requests from trusted entities, a differential privacy optimizer is configured to train a differential privacy model. The data that the trusted entity expects to acquire is processed into privacy data and sent to the trusted entity. Based on the processing logs of the privacy data by the trusted entity, the parameters of the differential privacy optimizer are adjusted. When the trusted entity requests data again, the differential privacy optimizer is used to retrain the differential privacy model. Differential privacy protection is independent of the data analyst's prior knowledge, resisting all attacks relying on auxiliary information. It provides privacy protection throughout the entire data processing process, ensuring that individual data is not identified, effectively avoiding the vulnerability of data de-identification and the privacy risks of data sharing. It is suitable for large-scale dataset sharing scenarios, improving the credibility and security of data sharing.
[0004] This invention is achieved through the following technical solution:
[0005] A method for sharing cloud-based terminal data with a platform based on differential privacy includes:
[0006] Collect operational data from user terminals and preprocess the operational data on the user terminal side to obtain preprocessed operational data; based on the operational status of the user terminal, send the preprocessed operational data to the cloud platform;
[0007] The preprocessed runtime data received by the cloud platform is filtered to obtain usable runtime data, and the usable runtime data is stored in the storage space corresponding to the cloud platform; based on the data acquisition request from the trusted entity, a corresponding differential privacy optimizer is configured to train the differential privacy model; based on the trained differential privacy model, the data that the trusted entity expects to acquire is processed into privacy data;
[0008] Based on the data transmission link between the cloud platform and the trusted entity, the privacy data is sent to the trusted entity; based on the processing log of the privacy data by the trusted entity, the parameters of the differential privacy optimizer are adjusted, so that when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
[0009] Optionally, the user terminal's operational data is collected, and the operational data is preprocessed on the user terminal side to obtain preprocessed operational data; based on the user terminal's operational status, the preprocessed operational data is sent to the cloud platform, including:
[0010] The user terminal is monitored to obtain the running thread log of the user terminal; based on the running thread log, the running data generation attribute information of all currently triggered task threads of the user terminal in the future time period is predicted; wherein, the running data generation attribute information includes running data generation rate information and running data generation port information;
[0011] Based on the operational data, attribute information is generated, and operational data generated by the user terminal in future time periods is collected. Based on the historical interaction data template information between the user terminal and the cloud platform, the operational data is preprocessed by data format conversion and data compression on the user terminal side to obtain preprocessed operational data.
[0012] Based on the running thread logs, determine whether the user terminal has completed the calculation and processing of all task threads; if so, send all currently generated preprocessed running data to the cloud platform; if not, based on the running thread logs, predict the time node when the user terminal will complete all task threads, and then send all preprocessed running data generated up to the time node to the cloud platform.
[0013] Optionally, sending the preprocessed runtime data to the cloud platform further includes:
[0014] The process of sending preprocessed runtime data to the cloud platform may experience data upload delays, causing the preprocessed runtime data to accumulate. Therefore, based on changes in the preprocessed runtime data, control is implemented to determine whether to clear the accumulated preprocessed runtime data, including:
[0015] Based on the data returned from the cloud platform, calculate the data upload latency T:
[0016]
[0017] In the formula, t(S) 16 ) indicates that S 16 The moment when data is sent to the cloud platform; S 16 H represents the hexadecimal form of the most recent preprocessing run data; 16 This represents the hexadecimal form of the data received from the cloud platform; t(H 16 =S 16 This indicates the received data returned from the cloud platform and S. 16 The moment when they are equal;
[0018] Based on the data upload delay and the generation status of the preprocessing data, the data volume D under the maximum accumulation state is obtained:
[0019]
[0020] In the formula, m represents the amount of data cleaned since the last cleanup; S(k) 16 T represents the hexadecimal form of the preprocessed runtime data sent to the cloud platform for the kth time; m d[S(k)] represents the total time for sending m preprocessed data runs to the cloud platform; 16 [] indicates the amount of preprocessed runtime data sent to the cloud platform in the k-th iteration; This means taking the values of k from 1 to m and substituting them into the parentheses to get the maximum value inside the parentheses;
[0021] Based on the changes in the preprocessing data and the amount of data at its maximum accumulation state, control the clearing of accumulated data and control the remaining amount of data Δd after clearing:
[0022]
[0023] In the formula, S(ka) 16 This represents the hexadecimal form of the preprocessed runtime data sent to the cloud platform for the kath iteration; 'a' represents an integer variable. This means that the value of a is gradually increased from 1 until it is found that S(k) is not satisfied. 16 -S(ka) 16When a = 0, the value of a is n, and the value of the whole expression is n; This indicates the expression within the parentheses. The minimum value between 1 and 2.
[0024] Optionally, the preprocessed runtime data received by the cloud platform is filtered to obtain usable runtime data, and the usable runtime data is stored in the storage space corresponding to the cloud platform; based on the data acquisition request from the trusted entity, a corresponding differential privacy optimizer is configured to train the differential privacy model; based on the trained differential privacy model, the data that the trusted entity expects to acquire is processed into privacy data, including:
[0025] The preprocessed running data received by the cloud platform is subjected to invalid and garbled data component identification and removal to obtain usable running data; based on the quantity of usable running data and the capacity of all available storage spaces under the cloud platform, the usable running data is stored in the corresponding available storage space;
[0026] The data acquisition request from the trusted entity is parsed and processed to obtain the data privacy protection strength feature information of the trusted entity; based on the data privacy protection strength feature information, the transformation operation parameters corresponding to the data privacy transformation are determined; wherein, the transformation operation parameters include any one or at least two of the noise ratio, clipping ratio and learning efficiency corresponding to the data privacy transformation; and then, based on the transformation operation parameters, a corresponding differential privacy optimizer is configured to train the differential privacy model.
[0027] Based on the trained differential privacy model, the data that the trusted entity expects to obtain is subjected to noise mixing and / or data fragment editing and disordering, thereby processing the data that the trusted entity expects to obtain into privacy data.
[0028] Optionally, based on the data transmission link between the cloud platform and the trusted entity, the privacy data is sent to the trusted entity; based on the trusted entity's processing log of the privacy data, the parameters of the differential privacy optimizer are adjusted, so that when the trusted entity requests data again, the differential privacy model is retrained using the parameter-adjusted differential privacy optimizer, including:
[0029] Based on the bandwidth characteristics of the data transmission link between the cloud platform and the trusted entity, the privacy data is divided into a privacy data block queue and then sent to the trusted entity.
[0030] Based on the trusted entity's processing logs of the privacy data, the type of restoration and recovery operation process of the privacy data by the trusted entity is determined. Based on this, the parameters of the differential privacy optimizer are adjusted to include the noise ratio and / or clipping ratio, and the correspondence between the differential privacy optimizer and the trusted entity is determined. Thus, when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
[0031] Furthermore, this invention also discloses a cloud-based terminal data and platform sharing system based on differential privacy, comprising:
[0032] The data collection and preprocessing module is used to collect the operating data of the user terminal and preprocess the operating data on the user terminal side to obtain preprocessed operating data.
[0033] The data sending module is used to send the preprocessed running data to the cloud platform based on the running status of the user terminal;
[0034] The data filtering and storage module is used to filter the preprocessed running data received by the cloud platform, obtain usable running data, and store the usable running data in the storage space corresponding to the cloud platform.
[0035] The data differential privacy processing module is used to configure a corresponding differential privacy optimizer based on the data acquisition request from a trusted entity, thereby training a differential privacy model; and based on the trained differential privacy model, to process the data that the trusted entity expects to acquire into privacy data.
[0036] A privacy data sharing module is used to send the privacy data to the trusted entity based on the data transmission link between the cloud platform and the trusted entity;
[0037] The differential privacy model adjustment module is used to adjust the parameters of the differential privacy optimizer based on the processing log of the privacy data by the trusted entity, so that when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
[0038] Optionally, the data collection and preprocessing module is used to collect operational data from the user terminal and preprocess the operational data on the user terminal side to obtain preprocessed operational data, including:
[0039] The user terminal is monitored to obtain the running thread log of the user terminal; based on the running thread log, the running data generation attribute information of all currently triggered task threads of the user terminal in the future time period is predicted; wherein, the running data generation attribute information includes running data generation rate information and running data generation port information;
[0040] Based on the operational data, attribute information is generated, and operational data generated by the user terminal in future time periods is collected. Based on the historical interaction data template information between the user terminal and the cloud platform, the operational data is preprocessed by data format conversion and data compression on the user terminal side to obtain preprocessed operational data.
[0041] The data sending module is used to send the preprocessed running data to the cloud platform based on the operating status of the user terminal, including:
[0042] Based on the running thread logs, determine whether the user terminal has completed the calculation and processing of all task threads; if so, send all currently generated preprocessed running data to the cloud platform; if not, based on the running thread logs, predict the time node when the user terminal will complete all task threads, and then send all preprocessed running data generated up to the time node to the cloud platform.
[0043] Optionally, the data filtering and storage module is used to filter the preprocessed runtime data received by the cloud platform to obtain usable runtime data, and store the usable runtime data in the storage space corresponding to the cloud platform, including:
[0044] The preprocessed running data received by the cloud platform is subjected to invalid and garbled data component identification and removal to obtain usable running data; based on the quantity of usable running data and the capacity of all available storage spaces under the cloud platform, the usable running data is stored in the corresponding available storage space;
[0045] The differential privacy processing module is used to configure a corresponding differential privacy optimizer based on data acquisition requests from trusted entities, thereby training a differential privacy model; based on the trained differential privacy model, it processes the data that the trusted entity expects to acquire into privacy data, including:
[0046] The data acquisition request from the trusted entity is parsed and processed to obtain the data privacy protection strength feature information of the trusted entity; based on the data privacy protection strength feature information, the transformation operation parameters corresponding to the data privacy transformation are determined; wherein, the transformation operation parameters include any one or at least two of the noise ratio, clipping ratio and learning efficiency corresponding to the data privacy transformation; and then, based on the transformation operation parameters, a corresponding differential privacy optimizer is configured to train the differential privacy model.
[0047] Based on the trained differential privacy model, the data that the trusted entity expects to obtain is subjected to noise mixing and / or data fragment editing and disordering, thereby processing the data that the trusted entity expects to obtain into privacy data.
[0048] Optionally, the privacy data sharing module is used to send the privacy data to the trusted entity based on the data transmission link between the cloud platform and the trusted entity, including:
[0049] Based on the bandwidth characteristics of the data transmission link between the cloud platform and the trusted entity, the privacy data is divided into a privacy data block queue and then sent to the trusted entity.
[0050] The differential privacy model adjustment module is used to adjust the parameters of the differential privacy optimizer based on the trusted entity's processing logs of the privacy data. This allows the differential privacy model to be retrained using the adjusted differential privacy optimizer when the trusted entity requests data again. The module includes:
[0051] Based on the trusted entity's processing logs of the privacy data, the type of restoration and recovery operation process of the privacy data by the trusted entity is determined. Based on this, the parameters of the differential privacy optimizer are adjusted to include the noise ratio and / or clipping ratio, and the correspondence between the differential privacy optimizer and the trusted entity is determined. Thus, when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
[0052] Furthermore, the present invention also discloses a computer-readable storage medium storing at least one executable instruction, which, when executed on an electronic device, causes the electronic device to perform the above-disclosed method for sharing cloud terminal data with a platform based on differential privacy.
[0053] Compared with the prior art, the present invention has the following beneficial effects:
[0054] This application provides a method and system for sharing cloud terminal data with a platform based on differential privacy. The system collects and preprocesses operational data from user terminals to obtain preprocessed operational data. This preprocessed data is then sent to the cloud platform based on the user terminal's operational status, ensuring comprehensive collection of user terminal data and the richness of data sources for the cloud platform. The system filters usable operational data from the preprocessed operational data received from the cloud platform and stores it in the cloud platform's storage space. Furthermore, based on data acquisition requests from trusted entities, a differential privacy optimizer is configured to train a differential privacy model. The data that the trusted entity expects to acquire is processed into privacy data and sent to the trusted entity. Based on the trusted entity's processing logs of the privacy data, the parameters of the differential privacy optimizer are adjusted. When the trusted entity requests data again, the differential privacy optimizer is used to retrain the differential privacy model. Differential privacy protection is independent of the data analyst's prior knowledge, resisting all attacks relying on auxiliary information. It provides privacy protection throughout the entire data processing process, ensuring that individual data is not identified, effectively mitigating the vulnerability of data de-identification and the privacy risks of data sharing. It is suitable for large-scale dataset sharing scenarios, improving the credibility and security of data sharing. Attached Figure Description
[0055] 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 of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:
[0056] Figure 1 This is a flowchart illustrating the cloud terminal data and platform sharing method based on differential privacy provided by the present invention.
[0057] Figure 2 This is a schematic diagram of the structure of the cloud terminal data and platform sharing system based on differential privacy provided by the present invention. Detailed Implementation
[0058] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.
[0059] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, it should be noted that, for ease of description, only the parts relevant to this application are shown in the accompanying drawings, not the entire structure. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this application.
[0060] The terms “comprising” and “having”, and any variations thereof, used in this application are intended to cover non-exclusive inclusion. For example, a process, method, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus.
[0061] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0062] Please see Figure 1 As shown, an embodiment of this application provides a method for sharing cloud terminal data with a platform based on differential privacy. This method includes:
[0063] Collect operational data from user terminals and preprocess the operational data on the user terminal side to obtain preprocessed operational data; based on the operational status of the user terminal, send the preprocessed operational data to the cloud platform;
[0064] The preprocessed runtime data received by the cloud platform is filtered to obtain usable runtime data, and the usable runtime data is stored in the corresponding storage space of the cloud platform; based on the data acquisition request from the trusted entity, the corresponding differential privacy optimizer is configured to train the differential privacy model; based on the trained differential privacy model, the data that the trusted entity expects to acquire is processed into privacy data;
[0065] Based on the data transmission link between the cloud platform and the trusted entity, the privacy data is sent to the trusted entity; based on the trusted entity's processing log of the privacy data, the parameters of the differential privacy optimizer are adjusted, so that when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
[0066] The beneficial effects of the above embodiments are as follows: This cloud terminal data and platform sharing method based on differential privacy collects and preprocesses the operating data of user terminals to obtain preprocessed operating data, which is then sent to the cloud platform based on the operating status of the user terminals, achieving complete collection of user terminals and ensuring the richness of data sources for the cloud platform; the preprocessed operating data received from the cloud platform is filtered for usable operating data and stored in the cloud platform's storage space; based on the data acquisition requests of trusted entities, a differential privacy optimizer is configured to train a differential privacy model, processing the data that the trusted entity expects to acquire into privacy data and sending it to the trusted entity; based on the processing logs of the trusted entity on the privacy data, the parameters of the differential privacy optimizer are adjusted, and when the trusted entity requests data again, the differential privacy optimizer is used to retrain the differential privacy model. Differential privacy protection is independent of the data analyst's prior knowledge, can resist all attacks caused by auxiliary information, provides privacy protection throughout the data processing process, ensures that individual data is not identified, effectively avoids the vulnerability of data de-identification and the privacy risks of data sharing, is suitable for large-scale dataset sharing scenarios, and improves the credibility and security of data sharing.
[0067] In another embodiment, operational data from the user terminal is collected and preprocessed on the user terminal side to obtain preprocessed operational data; based on the operational status of the user terminal, the preprocessed operational data is sent to the cloud platform, including:
[0068] The user terminal is monitored to obtain the running thread log of the user terminal; based on the running thread log, the running data generation attribute information of all currently triggered task threads of the user terminal in the future time period is predicted; wherein, the running data generation attribute information includes running data generation rate information and running data generation port information;
[0069] Based on the operational data, attribute information is generated, and operational data generated by the user terminal in future time periods is collected. Based on the historical interaction data template information between the user terminal and the cloud platform, the operational data is preprocessed by data format conversion and data compression on the user terminal side to obtain preprocessed operational data.
[0070] Based on the running thread log, determine whether the user terminal has completed the calculation and processing of all task threads; if so, send all currently generated preprocessed running data to the cloud platform; if not, based on the running thread log, predict the time node when the user terminal will complete all task threads, and then send all preprocessed running data generated up to that time node to the cloud platform.
[0071] The beneficial effects of the above embodiments are that smart terminals such as IoT devices, smartphones, and tablets process user events and execute tasks during operation, generating corresponding operational data. To comprehensively collect the operational data generated by user terminals, the user terminal is monitored to obtain its operational thread log. This log comprehensively records all task threads to be processed by the user terminal, facilitating the subsequent collection of operational data generated by the user terminal one by one based on this log. Specifically, based on the operational thread log, the operational data generation attribute information of all currently triggered task threads of the user terminal in the future time period is predicted, thereby accurately determining the corresponding operational data generation rate and operational data port location during the processing of all task threads by the user terminal in the future time period. Furthermore, based on this operational data generation attribute information, the operational data generated by the user terminal in the future time period is collected at the corresponding operational data port location at a matching data collection rate, avoiding any omissions in the collection of operational data. Furthermore, based on historical interaction data templates between the user terminal and the cloud platform, the system performs data format conversion and compression preprocessing on the user terminal side to obtain preprocessed running data. This ensures that the format and content of the preprocessed running data are compatible with the cloud platform, eliminating the need for the cloud platform to preprocess running data from different user terminals and reducing its data processing workload. Additionally, based on the running thread logs, the system determines whether the user terminal has completed the computation of all task threads. If so, all currently generated preprocessed running data is sent to the cloud platform. If not, based on the running thread logs, the system predicts the time point at which the user terminal will complete all task threads and then sends all preprocessed running data generated up to that time point to the cloud platform. This ensures that all the running data the user terminal needs to send is completely transmitted to the cloud platform in a single data transmission, preventing data loss.
[0072] In another embodiment, sending the preprocessed runtime data to the cloud platform further includes:
[0073] The process of sending preprocessed runtime data to the cloud platform may experience data upload delays, causing the preprocessed runtime data to accumulate. Therefore, based on changes in the preprocessed runtime data, control is implemented to determine whether to clear the accumulated preprocessed runtime data, including:
[0074] Step S1: Using formula (1) below, obtain the data upload delay based on the data returned by the cloud platform.
[0075]
[0076] In the above formula (1), T represents the data upload delay duration; t(S 16 ) indicates that S 16 The moment when data is sent to the cloud platform; S 16 This represents the hexadecimal form of the most recent preprocessing run data; H 16 This represents the hexadecimal form of the data received from the cloud platform; t(H 16 =S 16 This indicates the received data returned from the cloud platform and S. 16 The moment when they are equal;
[0077] Step S2: Using the following formula (2), based on the data upload delay time and the generation status of the preprocessing running data, obtain the amount of data under the maximum accumulation state;
[0078]
[0079] In the above formula (2), D represents the amount of data in the maximum accumulation state; m represents the amount of data since the last cleanup (or the total number of preprocessing runs sent to the cloud platform from the first data upload to the present if no data cleanup has been performed); S(k) 16 T represents the hexadecimal form of the preprocessed runtime data sent to the cloud platform for the kth time; m d[S(k)] represents the total time for sending m preprocessed data runs to the cloud platform; 16 [] indicates the amount of preprocessed runtime data sent to the cloud platform in the k-th iteration; This means taking the values of k from 1 to m and substituting them into the parentheses to get the maximum value inside the parentheses;
[0080] Step S3: Using the following formula (3), based on the changes in the preprocessing running data and the amount of data under the maximum accumulation state, control the clearing of accumulated data and control the remaining amount of data after clearing.
[0081]
[0082] In formula (3) above, Δd represents the amount of data remaining after clearing the accumulated data; S(ka) 16 This represents the hexadecimal form of the preprocessed runtime data sent to the cloud platform for the kath iteration; 'a' represents an integer variable. This means that the value of a is gradually increased from 1 until it is found that S(k) is not satisfied. 16 -S(ka) 16 When a = 0, the value of a is n, and the value of the whole expression is also n; This indicates the expression within the parentheses. The minimum value directly equal to 1.
[0083] The beneficial effects of the above embodiments are as follows: using the above formula (1), the data upload delay time is obtained based on the data returned by the cloud platform, which makes it easier to know the current accumulation status of the data based on the data upload delay time, thereby better controlling the accumulation of data; then using the above formula (2), the amount of data under the maximum accumulation state is obtained based on the data upload delay time and the generation status of the preprocessing running data, and then the amount of data under the maximum accumulation state is obtained based on the maximum data amount, thereby influencing the subsequent control of the clearing of the accumulated data amount by expanding the accumulation threshold, ensuring that the system can complete the processing of accumulated data normally under high load; then using the above formula (3), the accumulated data is cleared and the remaining amount of data after clearing is controlled based on the changes in the preprocessing running data and the amount of data under the maximum accumulation state, thereby clearing part of the accumulated data when the data changes little, ensuring the speed of system operation.
[0084] In another embodiment, the preprocessed runtime data received by the cloud platform is filtered to obtain usable runtime data, and the usable runtime data is stored in the storage space corresponding to the cloud platform; based on the data acquisition request from the trusted entity, a corresponding differential privacy optimizer is configured to train a differential privacy model; based on the trained differential privacy model, the data that the trusted entity expects to acquire is processed into privacy data, including:
[0085] Invalid and garbled data components are identified and removed from the preprocessed running data received by the cloud platform to obtain usable running data; based on the quantity of usable running data and the capacity of all available storage spaces under the cloud platform, the usable running data is stored in the corresponding available storage space;
[0086] The data acquisition request from the trusted entity is parsed and processed to obtain the data privacy protection strength feature information of the trusted entity; based on the data privacy protection strength feature information, the transformation operation parameters corresponding to the data privacy transformation are determined; wherein, the transformation operation parameters include any one or at least two of the noise ratio, clipping ratio and learning efficiency corresponding to the data privacy transformation; and based on the transformation operation parameters, the corresponding differential privacy optimizer is configured to train the differential privacy model.
[0087] Based on the trained differential privacy model, the data that the trusted entity expects to obtain is subjected to noise mixing and / or data fragment editing and disordering, thereby processing the data that the trusted entity expects to obtain into privacy data.
[0088] The beneficial effects of the above embodiments are as follows: Invalid and garbled data components are identified and removed from the preprocessed running data received by the cloud platform to obtain usable running data. Based on the quantity of usable running data and the capacity of all available storage spaces under the cloud platform, the usable running data is stored in the corresponding available storage space. This allows for noise filtering of the preprocessed running data, fully utilizing the cloud platform's storage space and avoiding its invalid occupation. Furthermore, data acquisition requests from trusted entities are parsed to obtain the data privacy protection strength characteristics of the trusted entity. This determines any one or at least two of the noise ratio, clipping ratio, and learning efficiency corresponding to data privacy transformation. This allows for the quantification and calibration of subsequent data privacy transformation requirements, defining the target for training a differential privacy model using a differential privacy optimizer. Additionally, based on the trained differential privacy model, noise is added to the data that the trusted entity expects to acquire, and / or data fragments are clipped and disordered, thereby processing the data into privacy data. This enables data privacy transformation tailored to the needs of trusted entities.
[0089] In another embodiment, the privacy data is sent to the trusted entity based on the data transmission link between the cloud platform and the trusted entity; based on the trusted entity's processing log of the privacy data, the parameters of the differential privacy optimizer are adjusted, so that when the trusted entity requests data again, the differential privacy model is retrained using the parameter-adjusted differential privacy optimizer, including:
[0090] Based on the bandwidth characteristics of the data transmission link between the cloud platform and the trusted entity, the privacy data is divided into a privacy data block queue and then sent to the trusted entity.
[0091] Based on the trusted entity's processing logs of the privacy data, the type of restoration and recovery operation process of the privacy data is determined. Based on this, the parameters of the differential privacy optimizer are adjusted to determine the noise ratio and / or clipping ratio, and the correspondence between the differential privacy optimizer and the trusted entity is determined. Thus, when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
[0092] The beneficial effects of the above embodiments are as follows: Based on the bandwidth characteristics of the data transmission link between the cloud platform and the trusted entity, the privacy data is divided into a privacy data block queue and sent to the trusted entity. This improves the efficiency of shared transmission of privacy data from the cloud platform to the trusted entity and avoids congestion during the shared transmission process. Furthermore, based on the trusted entity's processing logs of the privacy data, the type of restoration and recovery operation process for the privacy data is determined. This allows for parameter adjustment of the noise ratio and / or clipping ratio of the differential privacy optimizer, and establishes the correspondence between the differential privacy optimizer and the trusted entity. Therefore, when the trusted entity requests data again, the differential privacy model is retrained using the parameter-adjusted differential privacy optimizer. This feedback training of the differential privacy model further improves the matching accuracy between the differential privacy model's data privacy transformation and the corresponding trusted entity.
[0093] Please see Figure 2 As shown, an embodiment of this application provides a cloud-based terminal data and platform sharing system based on differential privacy. This cloud-based terminal data and platform sharing system based on differential privacy includes:
[0094] The data collection and preprocessing module is used to collect the operating data of the user terminal and preprocess the operating data on the user terminal side to obtain preprocessed operating data.
[0095] The data sending module is used to send the preprocessed running data to the cloud platform based on the running status of the user terminal;
[0096] The data filtering and storage module is used to filter the preprocessed running data received by the cloud platform, obtain usable running data, and store the usable running data in the corresponding storage space of the cloud platform.
[0097] The differential privacy processing module is used to configure the corresponding differential privacy optimizer based on the data acquisition request from the trusted entity, thereby training the differential privacy model; based on the trained differential privacy model, the data that the trusted entity expects to acquire is processed into privacy data.
[0098] The privacy data sharing module is used to send the privacy data to the trusted entity based on the data transmission link between the cloud platform and the trusted entity;
[0099] The differential privacy model adjustment module is used to adjust the parameters of the differential privacy optimizer based on the processing log of the private data by the trusted entity, so that when the trusted entity requests data again, the differential privacy model can be retrained through the differential privacy optimizer with adjusted parameters.
[0100] The beneficial effects of the above embodiments are as follows: This cloud terminal data and platform sharing system based on differential privacy collects and preprocesses the operating data of user terminals to obtain preprocessed operating data, which is then sent to the cloud platform based on the operating status of the user terminals, achieving complete collection of user terminals and ensuring the richness of data sources for the cloud platform; the preprocessed operating data received from the cloud platform is filtered for usable operating data and stored in the cloud platform's storage space; based on the data acquisition requests of trusted entities, a differential privacy optimizer is configured to train a differential privacy model, processing the data that the trusted entity expects to acquire into privacy data and sending it to the trusted entity; based on the processing logs of the trusted entity on the privacy data, the parameters of the differential privacy optimizer are adjusted, and when the trusted entity requests data again, the differential privacy optimizer is used to retrain the differential privacy model. Differential privacy protection is independent of the data analyst's prior knowledge, can resist all attacks caused by auxiliary information, provides privacy protection throughout the data processing process, ensures that individual data is not identified, effectively avoids the vulnerability of data de-identification and the privacy risks of data sharing, is suitable for large-scale dataset sharing scenarios, and improves the credibility and security of data sharing.
[0101] In another embodiment, the data collection and preprocessing module is used to collect operational data from the user terminal and preprocess the operational data on the user terminal side to obtain preprocessed operational data, including:
[0102] The user terminal is monitored to obtain the running thread log of the user terminal; based on the running thread log, the running data generation attribute information of all currently triggered task threads of the user terminal in the future time period is predicted; wherein, the running data generation attribute information includes running data generation rate information and running data generation port information;
[0103] Based on the operational data, attribute information is generated, and operational data generated by the user terminal in future time periods is collected. Based on the historical interaction data template information between the user terminal and the cloud platform, the operational data is preprocessed by data format conversion and data compression on the user terminal side to obtain preprocessed operational data.
[0104] This data sending module is used to send the preprocessed runtime data to the cloud platform based on the user terminal's operating status, including:
[0105] Based on the running thread log, determine whether the user terminal has completed the calculation and processing of all task threads; if so, send all currently generated preprocessed running data to the cloud platform; if not, based on the running thread log, predict the time node when the user terminal will complete all task threads, and then send all preprocessed running data generated up to that time node to the cloud platform.
[0106] The beneficial effects of the above embodiments are that smart terminals such as IoT devices, smartphones, and tablets process user events and execute tasks during operation, generating corresponding operational data. To comprehensively collect the operational data generated by user terminals, the user terminal is monitored to obtain its operational thread log. This log comprehensively records all task threads to be processed by the user terminal, facilitating the subsequent collection of operational data generated by the user terminal one by one based on this log. Specifically, based on the operational thread log, the operational data generation attribute information of all currently triggered task threads of the user terminal in the future time period is predicted, thereby accurately determining the corresponding operational data generation rate and operational data port location during the processing of all task threads by the user terminal in the future time period. Furthermore, based on this operational data generation attribute information, the operational data generated by the user terminal in the future time period is collected at the corresponding operational data port location at a matching data collection rate, avoiding any omissions in the collection of operational data. Furthermore, based on historical interaction data templates between the user terminal and the cloud platform, the system performs data format conversion and compression preprocessing on the user terminal side to obtain preprocessed running data. This ensures that the format and content of the preprocessed running data are compatible with the cloud platform, eliminating the need for the cloud platform to preprocess running data from different user terminals and reducing its data processing workload. Additionally, based on the running thread logs, the system determines whether the user terminal has completed the computation of all task threads. If so, all currently generated preprocessed running data is sent to the cloud platform. If not, based on the running thread logs, the system predicts the time point at which the user terminal will complete all task threads and then sends all preprocessed running data generated up to that time point to the cloud platform. This ensures that all the running data the user terminal needs to send is completely transmitted to the cloud platform in a single data transmission, preventing data loss.
[0107] In another embodiment, the data filtering and storage module is used to filter the preprocessed runtime data received by the cloud platform to obtain usable runtime data, and store the usable runtime data in the storage space corresponding to the cloud platform, including:
[0108] Invalid and garbled data components are identified and removed from the preprocessed running data received by the cloud platform to obtain usable running data; based on the quantity of usable running data and the capacity of all available storage spaces under the cloud platform, the usable running data is stored in the corresponding available storage space;
[0109] This differential privacy processing module is used to configure a corresponding differential privacy optimizer based on data acquisition requests from trusted entities, thereby training a differential privacy model; based on the trained differential privacy model, it processes the data that the trusted entity expects to acquire into privacy data, including:
[0110] The data acquisition request from the trusted entity is parsed and processed to obtain the data privacy protection strength feature information of the trusted entity; based on the data privacy protection strength feature information, the transformation operation parameters corresponding to the data privacy transformation are determined; wherein, the transformation operation parameters include any one or at least two of the noise ratio, clipping ratio and learning efficiency corresponding to the data privacy transformation; and based on the transformation operation parameters, the corresponding differential privacy optimizer is configured to train the differential privacy model.
[0111] Based on the trained differential privacy model, the data that the trusted entity expects to obtain is subjected to noise mixing and / or data fragment editing and disordering, thereby processing the data that the trusted entity expects to obtain into privacy data.
[0112] The beneficial effects of the above embodiments are as follows: Invalid and garbled data components are identified and removed from the preprocessed running data received by the cloud platform to obtain usable running data. Based on the quantity of usable running data and the capacity of all available storage spaces under the cloud platform, the usable running data is stored in the corresponding available storage space. This allows for noise filtering of the preprocessed running data, fully utilizing the cloud platform's storage space and avoiding its invalid occupation. Furthermore, data acquisition requests from trusted entities are parsed to obtain the data privacy protection strength characteristics of the trusted entity. This determines any one or at least two of the noise ratio, clipping ratio, and learning efficiency corresponding to data privacy transformation. This allows for the quantification and calibration of subsequent data privacy transformation requirements, defining the target for training a differential privacy model using a differential privacy optimizer. Additionally, based on the trained differential privacy model, noise is added to the data that the trusted entity expects to acquire, and / or data fragments are clipped and disordered, thereby processing the data into privacy data. This enables data privacy transformation tailored to the needs of trusted entities.
[0113] In another embodiment, the privacy data sharing module is used to send the privacy data to the trusted entity based on the data transmission link between the cloud platform and the trusted entity, including:
[0114] Based on the bandwidth characteristics of the data transmission link between the cloud platform and the trusted entity, the privacy data is divided into a privacy data block queue and then sent to the trusted entity.
[0115] The differential privacy model tuning module is used to adjust the parameters of the differential privacy optimizer based on the trusted entity's processing logs of the privacy data. This allows the differential privacy model to be retrained using the adjusted differential privacy optimizer when the trusted entity requests data again. The module includes:
[0116] Based on the trusted entity's processing logs of the privacy data, the type of restoration and recovery operation process of the privacy data is determined. Based on this, the parameters of the differential privacy optimizer are adjusted to determine the noise ratio and / or clipping ratio, and the correspondence between the differential privacy optimizer and the trusted entity is determined. Thus, when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
[0117] The beneficial effects of the above embodiments are as follows: Based on the bandwidth characteristics of the data transmission link between the cloud platform and the trusted entity, the privacy data is divided into a privacy data block queue and sent to the trusted entity. This improves the efficiency of shared transmission of privacy data from the cloud platform to the trusted entity and avoids congestion during the shared transmission process. Furthermore, based on the trusted entity's processing logs of the privacy data, the type of restoration and recovery operation process for the privacy data is determined. This allows for parameter adjustment of the noise ratio and / or clipping ratio of the differential privacy optimizer, and establishes the correspondence between the differential privacy optimizer and the trusted entity. Therefore, when the trusted entity requests data again, the differential privacy model is retrained using the parameter-adjusted differential privacy optimizer. This feedback training of the differential privacy model further improves the matching accuracy between the differential privacy model's data privacy transformation and the corresponding trusted entity.
[0118] In summary, this cloud-based terminal data sharing method and system based on differential privacy collects and preprocesses user terminal operational data to obtain preprocessed operational data. This preprocessed operational data is then sent to the cloud platform based on the user terminal's operational status, ensuring comprehensive collection of user terminal data and the richness of data sources for the cloud platform. The system filters usable operational data from the preprocessed operational data received from the cloud platform and stores it in the cloud platform's storage space. Furthermore, based on trusted entities' data acquisition requests, a differential privacy optimizer is configured to train a differential privacy model. The data that the trusted entity expects to acquire is processed into privacy data and sent to the trusted entity. Based on the trusted entity's processing logs of the privacy data, the parameters of the differential privacy optimizer are adjusted. When the trusted entity requests data again, the differential privacy optimizer is used to retrain the differential privacy model. Differential privacy protection is independent of the data analyst's prior knowledge, resisting all attacks relying on auxiliary information. It provides privacy protection throughout the entire data processing process, ensuring that individual data is not identified, effectively mitigating the vulnerability of data de-identification and the privacy risks of data sharing. It is suitable for large-scale dataset sharing scenarios, improving the credibility and security of data sharing.
[0119] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. Although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the invention itself. Various changes in form and detail may be made without departing from the spirit and scope of the invention as defined in the appended claims.
Claims
1. A method for sharing cloud terminal data with a platform based on differential privacy, characterized in that, Includes the following steps: Collect operational data from user terminals and preprocess the operational data on the user terminal side to obtain preprocessed operational data; based on the operational status of the user terminal, send the preprocessed operational data to the cloud platform; The process of sending preprocessed runtime data to the cloud platform may experience data upload delays, causing the preprocessed runtime data to accumulate. Therefore, based on changes in the preprocessed runtime data, control is implemented to determine whether to clear the accumulated preprocessed runtime data, specifically including: Based on the data returned from the cloud platform, calculate the data upload latency T: In the formula, Indicates will The moment when data is sent to the cloud platform; This represents the hexadecimal form of the most recent preprocessing run data; This represents the hexadecimal form of the data received from the cloud platform. This indicates the received data returned from the cloud platform and The moment when they are equal; Based on the data upload delay and the generation status of the preprocessed data, the data volume D under the maximum accumulation state is obtained: In the formula, m represents the amount of data cleaned since the last cleanup. If no data has been cleaned, it represents the total number of times the preprocessed data has been sent to the cloud platform since the first data upload. Indicates the first The hexadecimal form of the pre-processed running data sent to the cloud platform; express The total time for sending preprocessed data to the cloud platform; Indicates the first The amount of pre-processed runtime data sent to the cloud platform each time; Indicates will The value ranges from 1 to Substitute the value into the parentheses to get the maximum value inside the parentheses; Based on the changes in the preprocessing data and the amount of data at its maximum accumulation state, control the clearing of accumulated data and control the remaining amount of data after clearing. : In the formula, Indicates the first The hexadecimal form of the pre-processed running data sent to the cloud platform; Represents integer variables; Indicates will The value starts from 1 and gradually increases until it is found that the condition is not met. time Value The overall numerical value of the expression is ; This indicates the expression within the parentheses. The minimum value between 1 and 2; The preprocessed runtime data received by the cloud platform is filtered to obtain usable runtime data, and the usable runtime data is stored in the storage space corresponding to the cloud platform; based on the data acquisition request from the trusted entity, a corresponding differential privacy optimizer is configured to train the differential privacy model; based on the trained differential privacy model, the data that the trusted entity expects to acquire is processed into privacy data; Based on the data transmission link between the cloud platform and the trusted entity, the privacy data is sent to the trusted entity; based on the processing log of the privacy data by the trusted entity, the parameters of the differential privacy optimizer are adjusted, so that when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
2. The cloud terminal data and platform sharing method based on differential privacy according to claim 1, characterized in that, Collect the operating data of the user terminal, and preprocess the operating data on the user terminal side to obtain preprocessed operating data; Based on the operating status of the user terminal, the preprocessed operating data is sent to the cloud platform, specifically including: The user terminal is monitored to obtain the running thread log of the user terminal; based on the running thread log, the running data generation attribute information of all currently triggered task threads of the user terminal in the future time period is predicted; wherein, the running data generation attribute information includes running data generation rate information and running data generation port information; Based on the operational data, attribute information is generated, and operational data generated by the user terminal in future time periods is collected. Based on the historical interaction data template information between the user terminal and the cloud platform, the operational data is preprocessed by data format conversion and data compression on the user terminal side to obtain preprocessed operational data. Based on the running thread logs, determine whether the user terminal has completed the calculation and processing of all task threads; if so, send all currently generated preprocessed running data to the cloud platform; if not, based on the running thread logs, predict the time node when the user terminal will complete all task threads, and then send all preprocessed running data generated up to the time node to the cloud platform.
3. The cloud terminal data and platform sharing method based on differential privacy according to claim 1, characterized in that, The preprocessed running data received by the cloud platform is filtered to obtain usable running data, and the usable running data is stored in the storage space corresponding to the cloud platform. Based on data acquisition requests from trusted entities, configure the corresponding differential privacy optimizer to train the differential privacy model; Based on the trained differential privacy model, the data that the trusted entity expects to obtain is processed into privacy data, specifically including: The preprocessed running data received by the cloud platform is subjected to invalid and garbled data component identification and removal to obtain usable running data; based on the quantity of usable running data and the capacity of all available storage spaces under the cloud platform, the usable running data is stored in the corresponding available storage space; The data acquisition request from the trusted entity is parsed and processed to obtain the data privacy protection strength feature information of the trusted entity; based on the data privacy protection strength feature information, the transformation operation parameters corresponding to the data privacy transformation are determined; wherein, the transformation operation parameters include any one or at least two of the noise ratio, clipping ratio and learning efficiency corresponding to the data privacy transformation; and then, based on the transformation operation parameters, a corresponding differential privacy optimizer is configured to train the differential privacy model. Based on the trained differential privacy model, the data that the trusted entity expects to obtain is subjected to noise mixing and / or data fragment editing and disordering, thereby processing the data that the trusted entity expects to obtain into privacy data.
4. The cloud terminal data and platform sharing method based on differential privacy according to claim 1, characterized in that, Based on the data transmission link between the cloud platform and the trusted entity, the privacy data is sent to the trusted entity; based on the trusted entity's processing log of the privacy data, the parameters of the differential privacy optimizer are adjusted, so that when the trusted entity requests data again, the differential privacy model is retrained using the parameter-adjusted differential privacy optimizer, specifically including: Based on the bandwidth characteristics of the data transmission link between the cloud platform and the trusted entity, the privacy data is divided into a privacy data block queue and then sent to the trusted entity. Based on the trusted entity's processing logs of the privacy data, the type of restoration and recovery operation process of the privacy data by the trusted entity is determined. Based on this, the parameters of the differential privacy optimizer are adjusted to include the noise ratio and / or clipping ratio, and the correspondence between the differential privacy optimizer and the trusted entity is determined. Thus, when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
5. A cloud-based terminal data and platform sharing system based on differential privacy, comprising executing the cloud-based terminal data and platform sharing method based on differential privacy as described in any one of claims 1 to 4, characterized in that, The cloud-based terminal data and platform sharing system includes: The data collection and preprocessing module is used to collect the operating data of the user terminal and preprocess the operating data on the user terminal side to obtain preprocessed operating data. The data sending module is used to send the preprocessed running data to the cloud platform based on the running status of the user terminal; The data filtering and storage module is used to filter the preprocessed running data received by the cloud platform, obtain usable running data, and store the usable running data in the storage space corresponding to the cloud platform. The data differential privacy processing module is used to configure a corresponding differential privacy optimizer based on the data acquisition request from a trusted entity, thereby training a differential privacy model; and based on the trained differential privacy model, to process the data that the trusted entity expects to acquire into privacy data. A privacy data sharing module is used to send the privacy data to the trusted entity based on the data transmission link between the cloud platform and the trusted entity; The differential privacy model adjustment module is used to adjust the parameters of the differential privacy optimizer based on the processing log of the privacy data by the trusted entity, so that when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
6. The cloud terminal data and platform sharing system based on differential privacy according to claim 5, characterized in that, The data collection and preprocessing module is used to collect operational data from the user terminal and preprocess the operational data on the user terminal side to obtain preprocessed operational data, including: The user terminal is monitored to obtain the running thread log of the user terminal; based on the running thread log, the running data generation attribute information of all currently triggered task threads of the user terminal in the future time period is predicted; wherein, the running data generation attribute information includes running data generation rate information and running data generation port information; Based on the operational data, attribute information is generated, and operational data generated by the user terminal in future time periods is collected. Based on the historical interaction data template information between the user terminal and the cloud platform, the operational data is preprocessed by data format conversion and data compression on the user terminal side to obtain preprocessed operational data. The data sending module is used to send the preprocessed running data to the cloud platform based on the operating status of the user terminal, including: Based on the running thread logs, determine whether the user terminal has completed the calculation and processing of all task threads; if so, send all currently generated preprocessed running data to the cloud platform; if not, based on the running thread logs, predict the time node when the user terminal will complete all task threads, and then send all preprocessed running data generated up to the time node to the cloud platform.
7. The cloud terminal data and platform sharing system based on differential privacy according to claim 5, characterized in that, The data filtering and storage module is used to filter the preprocessed runtime data received by the cloud platform to obtain usable runtime data, and to store the usable runtime data in the storage space corresponding to the cloud platform, including: The preprocessed running data received by the cloud platform is subjected to invalid and garbled data component identification and removal to obtain usable running data; based on the quantity of usable running data and the capacity of all available storage spaces under the cloud platform, the usable running data is stored in the corresponding available storage space; The differential privacy processing module is used to configure a corresponding differential privacy optimizer based on data acquisition requests from trusted entities, thereby training a differential privacy model; based on the trained differential privacy model, it processes the data that the trusted entity expects to acquire into privacy data, including: The data acquisition request from the trusted entity is parsed and processed to obtain the data privacy protection strength feature information of the trusted entity; based on the data privacy protection strength feature information, the transformation operation parameters corresponding to the data privacy transformation are determined; wherein, the transformation operation parameters include any one or at least two of the noise ratio, clipping ratio and learning efficiency corresponding to the data privacy transformation; and then, based on the transformation operation parameters, a corresponding differential privacy optimizer is configured to train the differential privacy model. Based on the trained differential privacy model, the data that the trusted entity expects to obtain is subjected to noise mixing and / or data fragment editing and disordering, thereby processing the data that the trusted entity expects to obtain into privacy data.
8. The cloud terminal data and platform sharing system based on differential privacy according to claim 5, characterized in that, The privacy data sharing module is used to send the privacy data to the trusted entity based on the data transmission link between the cloud platform and the trusted entity, including: Based on the bandwidth characteristics of the data transmission link between the cloud platform and the trusted entity, the privacy data is divided into a privacy data block queue and then sent to the trusted entity. The differential privacy model adjustment module is used to adjust the parameters of the differential privacy optimizer based on the trusted entity's processing logs of the privacy data. This allows the differential privacy model to be retrained using the adjusted differential privacy optimizer when the trusted entity requests data again. The module includes: Based on the trusted entity's processing logs of the privacy data, the type of restoration and recovery operation process of the privacy data by the trusted entity is determined. Based on this, the parameters of the differential privacy optimizer are adjusted to include the noise ratio and / or clipping ratio, and the correspondence between the differential privacy optimizer and the trusted entity is determined. Thus, when the trusted entity requests data again, the differential privacy model is retrained through the differential privacy optimizer with adjusted parameters.
9. A computer-readable storage medium, characterized in that, The storage medium stores at least one executable instruction, which, when executed on an electronic device, causes the electronic device to perform the cloud terminal data and platform sharing method based on differential privacy as described in any one of claims 1 to 4.
Citation Information
Patent Citations
Multi-party information security calculation method and system based on differential privacy
CN117874821A
