Data privacy protection method, device and storage medium based on edge computing

Abstract

The present application discloses a data privacy protection method, device and storage medium based on edge computing. It relates to the field of information security technology. The method includes: in a federated machine learning mechanism, obtaining multiple training sub-models, wherein each training sub-model is obtained by training the initial model according to multiple original data by different terminal users, wherein the multiple original data are different terminal user data collected; associating each training sub-model according to a distributed hash table deployed on the edge node, so that the corresponding parameters between each training sub-model are interactively shared, and the shared model parameters are obtained; learning and training the target training sub-model in the target edge device according to the model parameters to obtain the target training model; detecting whether each original data has a data leakage risk according to the target training model. Through this application, the problem of high risk of data leakage in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the field of information security technology, and specifically, to a data privacy protection method, device and storage medium based on edge computing. Background Art

[0002] The federated learning framework has become one of the preferred directions for data federation and knowledge migration. Wide openness is an inherent feature of the cloud computing environment, which determines that the federated learning servers responsible for publishing tasks in the cloud and their associated training models and massive data face extremely high risks of privacy leakage, and also brings certain obstacles to the construction of system data security and privacy protection mechanisms.

[0003] With respect to the high risk of data leakage in related technologies, no effective solution has been proposed so far. Summary of the invention

[0004] The main purpose of this application is to provide a data privacy protection method, device and storage medium based on edge computing to solve the problem of high risk of data leakage in related technologies.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a data privacy protection method based on edge computing is provided. The method includes: in a federated machine learning mechanism, obtaining multiple training sub-models, wherein each training sub-model is obtained by training the initial model according to multiple original data by different terminal users, wherein the multiple original data are different terminal user data collected; associating each training sub-model according to a distributed hash table deployed on the edge node, so that the corresponding parameters between each training sub-model are interactively shared, and the shared model parameters are obtained, wherein the shared model parameters include the model parameters of each training sub-model; learning and training the target training sub-model in the target edge device according to the model parameters to obtain the target training model; detecting whether each original data has a data leakage risk according to the target training model.

[0006] Furthermore, before obtaining multiple training sub-models, the method also includes: processing original data including target label information among multiple original data to obtain target data; and using the target data to learn and train the initial model to obtain multiple training sub-models.

[0007] Furthermore, processing the original data including the target label information from the multiple original data to obtain the target data includes: filtering out the original data including the target label information from the multiple original data; and performing noise interference processing on the original data including the target label information in a federated machine learning mechanism to obtain the target data.

[0008] Furthermore, before associating each training sub-model according to the distributed hash table deployed on the edge node, the method also includes: determining retrieval information for the original data corresponding to each terminal user; constructing a distributed hash table according to the retrieval information, wherein the original data corresponding to each terminal user corresponds one-to-one to each sub-table in the distributed hash table; and storing the distributed hash table in the edge node.

[0009] Furthermore, before learning and training the target training sub-model in the target edge device according to the model parameters to obtain the target training model, the method also includes: searching the distributed hash table according to the target keyword to obtain the target original data; and sending the model parameters to the target edge device corresponding to the target original data.

[0010] Furthermore, detecting whether each original data has a data leakage risk according to the target training model includes: extracting target features in each original data, wherein the target features include at least: a target field corresponding to the attribute of the original data, and a target dictionary value corresponding to the target field; and judging whether the original data has a data leakage risk through the target features.

[0011] Furthermore, after detecting whether each original data has a data leakage risk according to the target training model, the method also includes: determining a sharing control strategy according to the target training model, wherein the sharing control strategy is used to constrain the flow of the original data.

[0012] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a data privacy protection device based on edge computing is provided. The device includes: an acquisition unit, which is used to acquire multiple training sub-models in a federated machine learning mechanism, wherein each training sub-model is obtained by training the initial model according to multiple original data by different terminal users, wherein the multiple original data are different terminal user data collected; a sharing unit, which is used to associate each training sub-model according to a distributed hash table deployed on the edge node, so that the corresponding parameters between each training sub-model are interactively shared, and the shared model parameters are obtained, wherein the shared model parameters include the model parameters of each training sub-model; a first training unit, which is used to learn and train the target training sub-model in the target edge device according to the model parameters to obtain the target training model; a detection unit, which is used to detect whether each original data has a data leakage risk according to the target training model.

[0013] Furthermore, the device also includes: a processing unit, which is used to process the original data including target label information in the multiple original data before obtaining multiple training sub-models to obtain target data; a second training unit, which is used to use the target data to learn and train the initial model to obtain multiple training sub-models.

[0014] Furthermore, the processing unit includes: a screening module, used to screen out original data including target label information from multiple original data; an interference module, used to perform noise interference processing on the original data including target label information in a federated machine learning mechanism to obtain target data.

[0015] Furthermore, the device also includes: a first determination unit, used to determine the retrieval information of the original data corresponding to each terminal user before associating each training sub-model according to the distributed hash table deployed on the edge node; a construction unit, used to construct a distributed hash table according to the retrieval information, wherein the original data corresponding to each terminal user corresponds one-to-one to each sub-table in the distributed hash table; and a storage unit, used to store the distributed hash table in the edge node.

[0016] Furthermore, the device also includes: a third training unit, used to learn and train the target training sub-model in the target edge device according to the model parameters, and retrieve the distributed hash table according to the target keyword before obtaining the target training model to obtain the target original data; a sending unit, used to send the model parameters to the target edge device corresponding to the target original data.

[0017] Furthermore, the detection unit includes: an extraction module for extracting target features from each original data, wherein the target features include at least: a target field corresponding to the attributes of the original data and a target dictionary value corresponding to the target field; and a judgment module for judging whether the original data has a data leakage risk through the target features.

[0018] Furthermore, the device also includes: a second determination unit, which is used to determine a sharing control strategy according to the target training model after detecting whether each original data has a data leakage risk according to the target training model, wherein the sharing control strategy is used to restrict the flow of the original data.

[0019] Through this application, the following steps are adopted: in a federated machine learning mechanism, multiple training sub-models are obtained, wherein each training sub-model is obtained by training the initial model according to multiple original data by different terminal users, wherein the multiple original data are different terminal user data collected; each training sub-model is associated according to a distributed hash table deployed on the edge node, so that the corresponding parameters between each training sub-model are interactively shared, and the shared model parameters are obtained, wherein the shared model parameters include the model parameters of each training sub-model; the target training sub-model is trained in the target edge device according to the model parameters to obtain the target training model; each original data is detected according to the target training model whether there is a risk of data leakage, thereby solving the problem of high risk of data leakage in related technologies. Each training sub-model is associated through a distributed hash table, so that the corresponding parameters between each training sub-model are interactively shared, the target training model is determined, and each original data is detected according to the target training model whether there is a risk of data leakage, thereby achieving the effect of reducing the risk of data leakage. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0021] Figure 1 is a flow chart of a data privacy protection method based on edge computing provided according to an embodiment of the present application;

[0022] Figure 2 It is a schematic diagram of the system architecture of the data privacy protection method based on edge computing provided in an embodiment of the present application;

[0023] Figure 3 It is a schematic diagram of a model update mechanism of a data privacy protection method based on edge computing provided in an embodiment of the present application;

[0024] Figure 4 It is a schematic diagram of model combination sharing of a data privacy protection method based on edge computing provided in an embodiment of the present application;

[0025] Figure 5 is a schematic diagram of a data privacy protection device based on edge computing provided according to an embodiment of the present application;

[0026] Figure 6 This is a schematic diagram of the network architecture of an edge computing-based data privacy protection electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0028] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0031] Gossip protocol: a protocol for information exchange between nodes or processes based on the way epidemics spread.

[0032] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0033] The present invention is described below in conjunction with preferred implementation steps. Figure 1 is a flow chart of a data privacy protection method based on edge computing provided in an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0034] Step S101, in the federated machine learning mechanism, obtain multiple training sub-models, wherein each training sub-model is obtained by training an initial model by different terminal users based on multiple original data, wherein the multiple original data are different terminal user data collected.

[0035] Specifically, in the federated machine learning mechanism, different participants download the initial model locally, and the initial model is trained using the participant data (corresponding to the original data in this application) to obtain multiple training sub-models, wherein different participants correspond to different end users in this application.

[0036] Optionally, in the edge computing-based data privacy protection method provided in an embodiment of the present application, before obtaining multiple training sub-models, the method also includes: processing the original data including target label information in the multiple original data to obtain target data; using the target data to learn and train the initial model to obtain multiple training sub-models.

[0037] Specifically, in the data privacy protection method based on edge computing provided in an embodiment of the present application, the original data including target label information in multiple original data are processed to obtain the target data, including: screening out the original data including the target label information from the multiple original data; performing noise interference processing on the original data including the target label information in a federated machine learning mechanism to obtain the target data.

[0038] For example, since the data of different participants may contain sensitive information and cannot be inferred by malicious participants, the federated machine learning mechanism uses the Laplace method to add noise perturbations to the local original data containing label information of sensitive information from different participants, that is, Participant c i The local model is updated through the updating process Sent to other participants for joint model construction, where k i (t) is the initial model, and τ is the perturbation parameter. By desensitizing the original data, the data security of the modeling is improved.

[0039] Step S102, associate each training sub-model according to the distributed hash table deployed on the edge node, so that the corresponding parameters between each training sub-model are interactively shared, and the shared model parameters are obtained, wherein the shared model parameters include the model parameters of each training sub-model.

[0040] Specifically, Figure 2 As shown, the data of participant 1, the data of participant 2, and the data of participant 3 are used to train sub-models respectively, and the model parameters are shared through the edge device.

[0041] Optionally, in the data privacy protection method based on edge computing provided in an embodiment of the present application, before associating each training sub-model according to the distributed hash table deployed on the edge node, the method also includes: determining retrieval information for the original data corresponding to each terminal user; constructing a distributed hash table according to the retrieval information, wherein the original data corresponding to each terminal user corresponds one-to-one to each sub-table in the distributed hash table; and storing the distributed hash table in the edge node.

[0042] Specifically, the data retrieval information of federated learning participants is constructed as a distributed hash table and registered in the edge node. It should be noted that before all learning tasks start, specific task participants are confirmed by searching the distributed hash table, which further ensures the accuracy of modeling learning data.

[0043] Step S103: Perform learning and training on the target training sub-model in the target edge device according to the model parameters to obtain the target training model.

[0044] Specifically, Figure 3 As shown, after the local participant data obtains the local model (corresponding to the training sub-model in this application), the model parameters of the training sub-model are shared with each other through the edge devices through the distributed hash table deployed on the edge node, and the model is aggregated in the target edge device to obtain the global model (corresponding to the target training model in this application).

[0045] Optionally, in the data privacy protection method based on edge computing provided in an embodiment of the present application, the target training sub-model is learned and trained in the target edge device according to the model parameters to obtain the target training model. The method also includes: searching the distributed hash table according to the target keyword to obtain the target original data; and sending the model parameters to the target edge device corresponding to the target original data.

[0046] Specifically, in the context of decentralized server participation, different participants in the present application will select a sub-table in the distributed hash table for model propagation to reduce differential attacks during the update process. Different participants will transmit their local models to the edge devices (corresponding to the target edge devices in the present application) involved in the data of other participants corresponding to the sub-table according to the Gossip protocol, and improve the model as training progresses. In the repeated execution of the update process, the global model will achieve the specified accuracy to prevent malicious participants from inferring the model provider information through the differences in adjacent rounds of updates, further ensuring the security of the modeling data.

[0047] Step S104: Detect whether each original data has a data leakage risk according to the target training model.

[0048] Specifically, by testing the target training model, it is verified based on the test results whether the target training model meets the detection requirements for detecting data leakage.

[0049] Optionally, in the edge computing-based data privacy protection method provided in an embodiment of the present application, detecting whether each original data has a data leakage risk according to the target training model includes: extracting target features in each original data, wherein the target features include at least: a target field corresponding to the attribute of the original data, and a target dictionary value corresponding to the target field; and judging whether the original data has a data leakage risk through the target features.

[0050] Specifically, the global model (corresponding to the target training model in this application) will have valid features in the original data, such as corresponding combinations of attribute fields and dictionary values, etc., and data sensitivity can be detected based on the valid features. It should be noted that edge devices participate in collaborative detection, and all edge terminals involved use the target training model to detect the content of the participants they transmit, thereby determining data leakage incidents and improving the accuracy of model training.

[0051] Optionally, in the edge computing-based data privacy protection method provided in an embodiment of the present application, after detecting whether each original data has a data leakage risk according to the target training model, the method also includes: determining a sharing control strategy according to the target training model, wherein the sharing control strategy is used to constrain the flow of the original data.

[0052] Specifically, the shared control strategy involves 4W factors W = (T, P, O, X), namely: time (When), space (Where), object (Who) and usage attribute (What), and the control constraints are ([C, Q], W γ , B), where [C, Q] is the request of the data user, W γ is the current 4W factor, B is the decision made by the model owner in response to the request, where the edge device node set E = {e1, e2, …, e n}, the minimum set m = {m1, m2, ..., m k}for:

[0053]

[0054] That is, according to the task allocation of the target training model, the edge device is applied for and the training model shared by the participants is received, and the data record of the corresponding model parameters is obtained, where the training sub-models distributed in each edge device can be expressed as: {(mo1, k(mo1)), (mo2, k(mo2)), …, (mo k ,k(mo k))}, the shared model construction process can be regarded as the main model recovery phase, that is, the polynomial solution process, expressed as {x1, y1}, {x2, y2}, …, {x k ,y k}Substitute it into the polynomial coefficient to solve:

[0055]

[0056] For example, Figure 4 is a schematic diagram of model combination sharing of a data privacy protection method based on edge computing provided in an embodiment of the present application, such as Figure 4 As shown, the FileA[0] and FileA[1] models determined by participant model provider A and the FileB[2] and FileB[3] models determined by participant model provider B share parameters between training sub-models through edge devices and update their respective training sub-models, thereby obtaining the target training models FileA[0]FileA[1]FileB[2] and FileA[1]FileB[2]FileB[3] in the task distribution model.

[0057] In summary, the data privacy protection method based on edge computing provided by the embodiment of the present application obtains multiple training sub-models in the federated machine learning mechanism, wherein each training sub-model is obtained by training the initial model according to multiple original data by different terminal users, wherein the multiple original data are different terminal user data collected; each training sub-model is associated according to the distributed hash table deployed on the edge node, so that the corresponding parameters between each training sub-model are interactively shared, and the shared model parameters are obtained, wherein the shared model parameters include the model parameters of each training sub-model; the target training sub-model is trained in the target edge device according to the model parameters to obtain the target training model; each original data is detected according to the target training model whether there is a risk of data leakage, thereby solving the problem of high risk of data leakage in the related technology. Each training sub-model is associated through a distributed hash table, so that the corresponding parameters between each training sub-model are interactively shared, the target training model is determined, and each original data is detected according to the target training model whether there is a risk of data leakage, thereby achieving the effect of reducing the risk of data leakage.

[0058] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0059] The embodiment of the present application also provides a data privacy protection device based on edge computing. It should be noted that the data privacy protection device based on edge computing in the embodiment of the present application can be used to execute the data privacy protection method based on edge computing provided in the embodiment of the present application. The following introduces the data privacy protection device based on edge computing provided in the embodiment of the present application.

[0060] Figure 5 Schematic diagram of a data privacy protection device based on edge computing according to an embodiment of the present application. Figure 5 As shown, the device includes: an acquisition unit 501, a sharing unit 502, a first training unit 503, and a detection unit 504.

[0061] Specifically, the acquisition unit 501 is used to acquire multiple training sub-models in the federated machine learning mechanism, wherein each training sub-model is obtained by training an initial model by different terminal users according to multiple original data, wherein the multiple original data are different terminal user data collected;

[0062] A sharing unit 502 is used to associate each training sub-model according to a distributed hash table deployed on an edge node, so that the corresponding parameters between each training sub-model are interactively shared, and the shared model parameters are obtained, wherein the shared model parameters include the model parameters of each training sub-model;

[0063] A first training unit 503 is used to perform learning and training on a target training sub-model in a target edge device according to the model parameters to obtain a target training model;

[0064] The detection unit 504 is used to detect whether each original data has a data leakage risk according to the target training model.

[0065] In summary, the data privacy protection device based on edge computing provided by the embodiment of the present application obtains multiple training sub-models in the federated machine learning mechanism through the acquisition unit 501, wherein each training sub-model is obtained by training the initial model according to multiple original data by different terminal users, wherein the multiple original data are different terminal user data collected; the sharing unit 502 associates each training sub-model according to the distributed hash table deployed on the edge node, so that the corresponding parameters between each training sub-model are interactively shared, and the shared model parameters are obtained, wherein the shared model parameters include the model parameters of each training sub-model; the first training unit 503 learns and trains the target training sub-model in the target edge device according to the model parameters to obtain the target training model; the detection unit 504 detects whether each original data has a data leakage risk according to the target training model, which solves the problem of high risk of data leakage in the related technology. Each training sub-model is associated through a distributed hash table, so that the corresponding parameters between each training sub-model are interactively shared, the target training model is determined, and each original data is detected according to the target training model to have a data leakage risk, thereby achieving the effect of reducing the risk of data leakage.

[0066] Optionally, in the edge computing-based data privacy protection device provided in an embodiment of the present application, the device also includes: a processing unit, used to process the original data including target label information in the multiple original data before obtaining multiple training sub-models to obtain target data; a second training unit, used to use the target data to learn and train the initial model to obtain multiple training sub-models.

[0067] Optionally, in the edge computing-based data privacy protection device provided in an embodiment of the present application, the processing unit includes: a screening module, used to screen out original data including target label information from multiple original data; an interference module, used to perform noise interference processing on the original data including target label information in a federated machine learning mechanism to obtain target data.

[0068] Optionally, in the data privacy protection device based on edge computing provided in an embodiment of the present application, the device also includes: a first determination unit, used to determine the retrieval information of the original data corresponding to each terminal user before associating each training sub-model according to the distributed hash table deployed on the edge node; a construction unit, used to construct a distributed hash table according to the retrieval information, wherein the original data corresponding to each terminal user corresponds one-to-one to each sub-table in the distributed hash table; and a storage unit, used to store the distributed hash table in the edge node.

[0069] Optionally, in the data privacy protection device based on edge computing provided in an embodiment of the present application, the device also includes: a third training unit, used to learn and train the target training sub-model in the target edge device according to the model parameters, and retrieve the distributed hash table according to the target keyword before obtaining the target training model to obtain the target original data; a sending unit, used to send the model parameters to the target edge device corresponding to the target original data.

[0070] Optionally, in the edge computing-based data privacy protection device provided in an embodiment of the present application, the detection unit includes: an extraction module for extracting target features in each original data, wherein the target features include at least: a target field corresponding to the attributes of the original data, and a target dictionary value corresponding to the target field; and a judgment module for judging whether the original data has a data leakage risk through the target features.

[0071] Optionally, in the edge computing-based data privacy protection device provided in an embodiment of the present application, the device also includes: a second determination unit, for determining a sharing control strategy according to the target training model after detecting whether each original data has a data leakage risk according to the target training model, wherein the sharing control strategy is used to constrain the flow of the original data.

[0072] The data privacy protection device based on edge computing includes a processor and a memory. The above-mentioned acquisition unit 501, sharing unit 502, first training unit 503, detection unit 504, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0073] The processor contains a kernel, which calls the corresponding program unit from the memory. One or more kernels can be set, and data privacy protection based on edge computing can be performed by adjusting kernel parameters.

[0074] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0075] An embodiment of the present invention provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, a data privacy protection method based on edge computing is implemented.

[0076] An embodiment of the present invention provides a processor, which is used to run a program, wherein a data privacy protection method based on edge computing is executed when the program is running.

[0077] like Figure 6As shown, an embodiment of the present invention provides an electronic device, the device includes a processor, a memory, and a program stored in the memory and executable on the processor, and the processor implements the following steps when executing the program: in a federated machine learning mechanism, a plurality of training sub-models are obtained, wherein each training sub-model is obtained by training an initial model according to a plurality of original data by different terminal users, wherein the plurality of original data are different terminal user data collected; each training sub-model is associated according to a distributed hash table deployed on an edge node, so that the corresponding parameters between each training sub-model are interactively shared, and the shared model parameters are obtained, wherein the shared model parameters include the model parameters of each training sub-model; the target training sub-model is learned and trained in the target edge device according to the model parameters to obtain a target training model; and each original data is detected according to the target training model whether there is a risk of data leakage.

[0078] When the processor executes the program, the following steps are also implemented: before obtaining multiple training sub-models, the original data including the target label information in the multiple original data are processed to obtain the target data; the target data is used to learn and train the initial model to obtain multiple training sub-models.

[0079] When the processor executes the program, the following steps are also implemented: filtering out original data including target label information from multiple original data; performing noise interference processing on the original data including target label information in the federated machine learning mechanism to obtain target data.

[0080] When the processor executes the program, the following steps are also implemented: before associating each training sub-model according to the distributed hash table deployed on the edge node, determining the retrieval information of the original data corresponding to each terminal user; constructing a distributed hash table according to the retrieval information, wherein the original data corresponding to each terminal user corresponds one-to-one to each sub-table in the distributed hash table; and storing the distributed hash table in the edge node.

[0081] When the processor executes the program, the following steps are also implemented: the target training sub-model is learned and trained in the target edge device according to the model parameters, and before obtaining the target training model, the distributed hash table is searched according to the target keyword to obtain the target original data; the model parameters are sent to the target edge device corresponding to the target original data.

[0082] When the processor executes the program, the following steps are also implemented: extracting target features from each raw data, wherein the target features include at least: target fields corresponding to the attributes of the raw data, and target dictionary values ​​corresponding to the target fields; and judging whether the raw data has a data leakage risk through the target features.

[0083] When the processor executes the program, the following steps are also implemented: after detecting whether each original data has a data leakage risk according to the target training model, determining a sharing control strategy according to the target training model, wherein the sharing control strategy is used to constrain the flow of the original data.

[0084] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0085] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that is initialized with the following method steps: in a federated machine learning mechanism, multiple training sub-models are obtained, wherein each training sub-model is obtained by training an initial model based on multiple original data by different terminal users, wherein the multiple original data are different terminal user data collected; each training sub-model is associated according to a distributed hash table deployed on an edge node, so that the corresponding parameters between each training sub-model are interactively shared, and the shared model parameters are obtained, wherein the shared model parameters include the model parameters of each training sub-model; the target training sub-model is learned and trained in the target edge device according to the model parameters to obtain a target training model; and each original data is detected according to the target training model to determine whether there is a risk of data leakage.

[0086] When executed on a data processing device, it is also suitable for executing an initialization program having the following method steps: before obtaining multiple training sub-models, processing the original data including target label information in multiple original data to obtain target data; using the target data to learn and train the initial model to obtain multiple training sub-models.

[0087] When executed on a data processing device, it is also suitable for executing an initialization program having the following method steps: screening out original data including target label information from multiple original data; performing noise interference processing on the original data including target label information in a federated machine learning mechanism to obtain target data.

[0088] When executed on a data processing device, it is also suitable for executing a program that is initialized with the following method steps: before associating each training sub-model according to a distributed hash table deployed on an edge node, determining retrieval information for the original data corresponding to each terminal user; constructing a distributed hash table according to the retrieval information, wherein the original data corresponding to each terminal user corresponds one-to-one to each sub-table in the distributed hash table; and storing the distributed hash table in the edge node.

[0089] When executed on a data processing device, it is also suitable for executing an initialization program having the following method steps: learning and training the target training sub-model in the target edge device according to the model parameters, and before obtaining the target training model, searching the distributed hash table according to the target keyword to obtain the target original data; sending the model parameters to the target edge device corresponding to the target original data.

[0090] When executed on a data processing device, it is also suitable for executing a program initialized with the following method steps: extracting target features in each original data, wherein the target features include at least: target fields corresponding to the attributes of the original data, and target dictionary values ​​corresponding to the target fields; judging whether the original data has a risk of data leakage through the target features.

[0091] When executed on a data processing device, it is also suitable for executing an initialization program having the following method steps: after detecting whether each original data has a data leakage risk according to a target training model, determining a sharing control strategy according to the target training model, wherein the sharing control strategy is used to constrain the flow of the original data.

[0092] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0093] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0094] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0095] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0096] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0097] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0098] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0099] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0100] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0101] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A data privacy protection method based on edge computing, characterized in that: include: In the federated machine learning mechanism, a plurality of training sub-models are obtained, wherein each training sub-model is obtained by training an initial model by a different terminal user according to a plurality of original data, wherein the plurality of original data are different terminal user data collected; Associating each training sub-model according to a distributed hash table deployed on an edge node, so that corresponding parameters between each training sub-model are interactively shared, and obtaining shared model parameters, wherein the shared model parameters include model parameters of each training sub-model; Performing learning and training on the target training sub-model in the target edge device according to the model parameters to obtain a target training model; Detecting whether each raw data has a data leakage risk according to the target training model; Before associating each training sub-model according to the distributed hash table deployed on the edge node, the method further includes: Determine retrieval information of the original data corresponding to each terminal user; Constructing a distributed hash table according to the search information, wherein the original data corresponding to each terminal user corresponds one-to-one to each sub-table in the distributed hash table; The distributed hash table is stored in the edge node.

2. The method according to claim 1, characterized in that: Before acquiring the multiple training sub-models, the method further includes: Processing the original data including the target tag information among the plurality of original data to obtain the target data; The target data is used to perform learning and training on the initial model to obtain the multiple training sub-models.

3. The method according to claim 2, characterized in that Processing the original data including the target tag information among the plurality of original data to obtain the target data includes: Filtering out original data including target tag information from the plurality of original data; In the federated machine learning mechanism, noise interference processing is performed on the original data including the target label information to obtain the target data.

4. The method according to claim 1, characterized in that Before learning and training the target training sub-model in the target edge device according to the model parameters to obtain the target training model, the method further includes: Search the distributed hash table according to the target keyword to obtain the target original data; The model parameters are sent to a target edge device corresponding to the target original data.

5. The method according to claim 1, characterized in that Detecting whether each raw data has a data leakage risk based on the target training model includes: Extracting target features from each of the original data, wherein the target features at least include: a target field corresponding to an attribute of the original data and a target dictionary value corresponding to the target field; It is determined whether the original data has a risk of data leakage based on the target features.

6. The method according to claim 1, characterized in that After detecting whether each raw data has a data leakage risk according to the target training model, the method further includes: A sharing control strategy is determined according to the target training model, wherein the sharing control strategy is used to constrain the flow of the original data.

7. A data privacy protection device based on edge computing, characterized in that: include: An acquisition unit, used for acquiring a plurality of training sub-models in a federated machine learning mechanism, wherein each training sub-model is obtained by training an initial model by a different terminal user according to a plurality of original data, wherein the plurality of original data are collected data of different terminal users; A sharing unit, configured to associate each training sub-model according to a distributed hash table deployed on an edge node, so that corresponding parameters between each training sub-model are interactively shared, and obtain shared model parameters, wherein the shared model parameters include model parameters of each training sub-model; A first training unit is used to perform learning and training on a target training sub-model in a target edge device according to the model parameters to obtain a target training model; A detection unit, used to detect whether each raw data has a data leakage risk according to the target training model; The device also includes: a first determination unit, used to determine the retrieval information of the original data corresponding to each terminal user before associating each training sub-model according to the distributed hash table deployed on the edge node; a construction unit, used to construct a distributed hash table according to the retrieval information, wherein the original data corresponding to each terminal user corresponds one-to-one to each sub-table in the distributed hash table; and a storage unit, used to store the distributed hash table in the edge node.

8. A computer-readable storage medium, characterized in that: The storage medium includes a stored program, wherein the program executes the method according to any one of claims 1 to 6.

9. An electronic device, characterized in that: include: The method comprises one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises a computer program or instructions, which implement the method according to any one of claims 1 to 6 when executed by a processor.

Images