Data processing method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses a data processing method and device, equipment and a storage medium. The method comprises the steps of creating a key pair, sending a first public key to N network elements in a network, obtaining N pieces of first encryption information, generating parameter correction information based on a first private key and the N pieces of first encryption information, the parameter correction information comprises the total gradient and total loss information of the N network elements, and sending the parameter correction information to each network element, so that each network element updates the parameters of the to-be-trained model of the network element based on the parameter correction information. Visibly, by collecting the gradient and loss information of each network element and returning the obtained total gradient and total loss information to each network element, joint training can be performed on the model of each network element without leaking the data of each network element, so that the prediction accuracy of the model of each network element is improved.

Description

Technical Field

[0001] This application relates to the field of computer technologies, and particularly to a data processing method, a data processing apparatus, a computer device, and a computer-readable storage medium. Background Art

[0002] With the progress of scientific and technological research, network technologies have developed rapidly. More and more services need to rely on the network during execution; for example, information dissemination, multimedia data transmission, remote control driving, etc. To ensure the execution of services, a risk prediction model is usually used to predict possible risks in the network (such as network quality fluctuations, network storms, etc.). It is found that due to the privacy of data of different network elements in the network, each network element can usually only train the risk prediction model through the data collected by itself, resulting in insufficient model training. Summary of the Invention

[0003] Embodiments of this application provide a data processing method, apparatus, device, computer-readable storage medium, and product, which can jointly train the models of each network element without leaking the data of each network element.

[0004] On the one hand, embodiments of this application provide a data processing method, including:

[0005] Create a key pair, and send a first public key to N network elements in the network, where the first public key is the public key in the key pair, and N is an integer greater than 1;

[0006] Obtain N first encrypted messages. The i-th first encrypted message is obtained by encrypting the gradient and loss information of the i-th network element using the first public key. The gradient and loss information of the i-th network element is calculated based on the first prediction data of this network element. The first prediction data is obtained by predicting the target sample parameters of the i-th network element through a model to be trained. The target sample parameters are the sample parameters obtained after aligning the sample parameters of the i-th network element with the sample parameters of N-1 network elements. i is a positive integer less than or equal to N;

[0007] Generate parameter correction information based on the first private key and N first encrypted messages. The parameter correction information includes the total gradient and total loss information of N network elements. The first private key is the private key in the key pair;

[0008] Send the parameter correction information to each network element, so that each network element updates the parameters of the model to be trained of this network element based on the parameter correction information.

[0009] In the embodiment of the present application, a key pair is created, and the first public key is sent to N network elements in the network, and N first encrypted messages are obtained. The i-th first encrypted message is obtained by encrypting the gradient and loss information of the i-th network element. Based on the first private key and the N first encrypted messages, parameter correction information is generated. The parameter correction information includes the total gradient and total loss information of the N network elements. The parameter correction information is sent to each network element so that each network element updates the parameters of the model to be trained based on the parameter correction information. It can be seen that by collecting the gradient and loss information of each network element through a third party (such as a network data analysis network element) and returning the obtained total gradient and total loss information to each network element, the models of each network element can be jointly trained without disclosing the data (network parameters) of each network element, thereby improving the prediction accuracy of the models of each network element.

[0010] On the one hand, the embodiment of the present application provides a data processing method, including:

[0011] Obtain a first public key and target sample parameters. The target sample parameters are sample parameters obtained by aligning the sample parameters of the application function network element with the sample parameters of N - 1 network elements in the network, where N is an integer greater than 1;

[0012] Predict the target sample parameters through the model to be trained to obtain first prediction data;

[0013] Generate a first encrypted message based on the first prediction data and the first public key, and return the first encrypted message to the provider of the first public key. The first encrypted message is obtained by encrypting the gradient and loss information of the application function network element. The gradient and loss information of the application function network element is calculated based on the first prediction data;

[0014] Obtain parameter correction information sent by the provider of the first public key. The parameter correction information includes the total gradient and total loss information of N network elements. The total gradient and total loss information of the N network elements are calculated based on the gradient and loss information of the application function network element;

[0015] Update the parameters of the model to be trained based on the parameter correction information to obtain a risk prediction model.

[0016] In the embodiment of the present application, a first public key and target sample parameters are obtained. The target sample parameters are predicted by a model to be trained to obtain first prediction data. Based on the first prediction data and the first public key, first encrypted information is generated and returned to the provider of the first public key. The first encrypted information is obtained by encrypting the gradient and loss information of the application function network element. The parameter correction information sent by the provider of the first public key is obtained. The parameter correction information includes the total gradient and total loss information of N network elements. The parameters of the model to be trained are updated based on the parameter correction information to obtain a risk prediction model. It can be seen that by providing the gradient and loss information to a third party (such as a network data analysis network element), the application function network element can obtain the total gradient and total loss information returned by the third party based on the gradient and loss information of each network element, and update the parameters of the model to be trained based on the total gradient and total loss information, so as to achieve the purpose of jointly training the models of each network element without disclosing the data of each network element.

[0017] On the one hand, the embodiment of the present application provides a data processing device, which includes:

[0018] A processing unit, configured to create a key pair and send a first public key to N network elements in the network through a sending unit. The first public key is the public key in the key pair, and N is an integer greater than 1;

[0019] An obtaining unit, configured to obtain N pieces of first encrypted information. The i-th first encrypted information is obtained by encrypting the gradient and loss information of the i-th network element using the first public key. The gradient and loss information of the i-th network element is calculated based on the first prediction data of the network element. The first prediction data is obtained by predicting the target sample parameters of the i-th network element through the model to be trained. The target sample parameters are the sample parameters obtained by aligning the sample parameters of the i-th network element with the sample parameters of N-1 network elements, and i is a positive integer less than or equal to N;

[0020] The processing unit is further configured to generate parameter correction information based on the first private key and N pieces of first encrypted information. The parameter correction information includes the total gradient and total loss information of N network elements. The first private key is the private key in the key pair;

[0021] The sending unit is further configured to send the parameter correction information to each network element, so that each network element updates the parameters of the model to be trained of the network element based on the parameter correction information.

[0022] In an implementation manner, the processing unit is further configured to:

[0023] Obtain N second encrypted messages. The i-th second encrypted message is obtained by the i-th network element encrypting the second prediction data of the network element using the first public key. The second prediction data is obtained by predicting the target network parameters of the i-th network element using the i-th sub-model. The target network parameters are the network parameters obtained by aligning the network parameters of the i-th network element with the network parameters of N - 1 network elements.

[0024] Based on the first private key and N second encrypted messages, determine the service parameters associated with the target service.

[0025] Send the service parameters to the i-th network element so that the i-th network element adjusts the service strategy of the target service based on the service parameters.

[0026] In one implementation, the processing unit is used to determine the service parameters associated with the target service based on the first private key and N second encrypted messages, specifically:

[0027] Decrypt the N second encrypted messages using the first private key to obtain N plaintext data.

[0028] Obtain M third encrypted messages. The third encrypted message is obtained by a network element in the network encrypting the network parameters of the network element using the second public key. The second public key is provided by the i-th network element, and M is an integer greater than 1.

[0029] Aggregate the M third encrypted messages to obtain an aggregation result corresponding to the M third encrypted messages.

[0030] Package the aggregation result and the N plaintext data to obtain the service parameters associated with the target service.

[0031] In one implementation, the processing unit is used to determine the service parameters associated with the target service based on the first private key and N second encrypted messages, specifically:

[0032] Decrypt the N second encrypted messages using the first private key to obtain N plaintext data.

[0033] Perform integration processing on the N plaintext data to obtain the service parameters associated with the target service. The integration processing includes at least one of the following: data format conversion, data aggregation, data intersection, data filtering, data analysis.

[0034] In one implementation, the processing unit is used to generate parameter correction information based on the first private key and N first encrypted messages, specifically:

[0035] Decrypt the N first encrypted messages using the first private key to obtain the gradients and loss information of N network elements.

[0036] Sum the gradients of N network elements to obtain the total gradient of the N network elements;

[0037] Sum the loss information of N network elements to obtain the total loss information of the N network elements;

[0038] Package the total gradient and total loss information of N network elements to obtain parameter correction information.

[0039] In one implementation, the processing unit is further configured to:

[0040] Obtain a service subscription request, where the service subscription request carries the identifiers of N network elements participating in collaborative training and the identity information of the sender of the service subscription request;

[0041] Verify the identity information;

[0042] If the identity information passes the verification, return response information to the sender of the service subscription request, where the response information is used to indicate that the service subscription request has passed.

[0043] In one implementation, the network elements in the network include user equipment, radio access network elements, and application function network elements;

[0044] The network parameters of the user equipment include at least one of the following: single-user signal-to-interference-plus-noise ratio, single-user received signal strength indication, single-user reference signal received power, single-user reference signal received quality, single-user delay, single-user data rate;

[0045] The network parameters of the radio access network elements include at least one of the following: signal-to-interference-plus-noise ratio, received signal strength indication, reference signal received power, reference signal received quality, delay, data rate, base station identifier;

[0046] The network parameters of the application function network elements are determined based on the target service.

[0047] On the one hand, an embodiment of the present application provides a data processing device, and the data processing device includes:

[0048] An acquisition unit, configured to acquire a first public key and target sample parameters, where the target sample parameters are sample parameters obtained by aligning the sample parameters of the application function network element with the sample parameters of N-1 network elements in the network, and N is an integer greater than 1;

[0049] A processing unit, configured to predict the target sample parameters through a to-be-trained model to obtain first prediction data;

[0050] and generating first encrypted information based on the first prediction data and the first public key, and returning the first encrypted information to the provider of the first public key through the sending unit, where the first encrypted information is obtained by encrypting the gradient and loss information of the application function network element, and the gradient and loss information of the application function network element is calculated based on the first prediction data;

[0051] The obtaining unit is further configured to obtain parameter correction information sent by the provider of the first public key, where the parameter correction information includes the total gradient and total loss information of N network elements, and the total gradient and total loss information of the N network elements are calculated based on the gradient and loss information of the application function network element;

[0052] The processing unit is further configured to update the parameters of the model to be trained based on the parameter correction information to obtain a risk prediction model.

[0053] In one implementation, the processing unit is further configured to:

[0054] Obtain target network parameters, where the target network parameters are network parameters obtained by aligning the network parameters of the application function network element with the network parameters of N - 1 network elements in the network;

[0055] Use the risk prediction model to predict the target network parameters to obtain second prediction data;

[0056] Encrypt the second prediction data with the first public key to obtain second encrypted information, and return the second encrypted information to the provider of the first public key;

[0057] Obtain service parameters sent by the provider of the first public key, where the service parameters are associated with the target service;

[0058] Adjust the service strategy of the target service based on the service parameters.

[0059] In one implementation, the service parameters include an aggregation result and N plaintext data, the N plaintext data are obtained by decrypting the first encrypted information of the N network elements, the aggregation result is obtained by aggregating M third encrypted information, and the jth third encrypted information is obtained by the jth network element encrypting the network parameters of the network element with the second public key provided by the application function network element, and j is a positive integer less than or equal to M;

[0060] The processing unit is configured to adjust the service strategy of the target service based on the service parameters, specifically:

[0061] Decrypt the aggregation result with the second private key corresponding to the second public key to obtain aggregation parameters;

[0062] Adjust the service strategy of the target service based on the aggregation parameters and the N plaintext data.

[0063] In one implementation, the target service is a remote teleoperation service, and the service parameters include the communication quality at a first moment; the processing unit is configured to adjust the service strategy of the target service based on the service parameters, specifically:

[0064] If the communication quality at the first moment is lower than the quality threshold, then reduce the driving speed of the remotely controlled object at a second moment, where the second moment is before the first moment.

[0065] In one implementation, the processing unit is configured to generate first encrypted information based on first prediction data and a first public key, specifically:

[0066] Obtain N - 1 intermediate data sent by N - 1 network elements; the i-th intermediate data is obtained by the i-th network element predicting the target sample parameters of the network element using the model to be trained in the network element, or is obtained by the i-th network element aggregating the target sample parameters of the network element;

[0067] Calculate the gradient and loss information based on the first prediction data and the N - 1 intermediate data;

[0068] Encrypt the gradient and loss information using the first public key to obtain the first encrypted information.

[0069] Correspondingly, the present application provides a computer device, which includes:

[0070] A memory in which a computer program is stored;

[0071] A processor for loading the computer program to implement the above data processing method.

[0072] Correspondingly, the present application provides a computer-readable storage medium storing a computer program, and the computer program is adapted to be loaded and executed by the processor to implement the above data processing method.

[0073] Correspondingly, the present application provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions to enable the computer device to implement the above data processing method.

[0074] In the embodiments of the present application, a third party creates a key pair and sends the first public key to N network elements in the network, and obtains N first encrypted messages. The i-th first encrypted message is obtained by encrypting the gradient and loss information of the i-th network element. Based on the first private key and the N first encrypted messages, parameter correction information is generated. The parameter correction information includes the total gradient and total loss information of the N network elements. The parameter correction information is sent to each network element so that each network element updates the parameters of the model to be trained of the network element based on the parameter correction information. The application function network element obtains the first public key and the target sample parameters, predicts the target sample parameters through the model to be trained, obtains the first prediction data, generates the first encrypted message based on the first prediction data and the first public key, and returns the first encrypted message to the provider (third party) of the first public key. The first encrypted message is obtained by encrypting the gradient and loss information of the application function network element. The parameter correction information sent by the provider of the first public key is obtained. The parameter correction information includes the total gradient and total loss information of the N network elements. The parameters of the model to be trained are updated based on the parameter correction information to obtain the risk prediction model. It can be seen that by collecting the gradient and loss information of each network element through a third party (such as a network data analysis network element) and returning the total gradient and total loss information obtained based on the gradient and loss information of each network element to each network element, the models of each network element can be jointly trained without disclosing the data (network parameters) of each network element, thereby improving the prediction accuracy of the models of each network element. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 It is a schematic diagram of a network architecture provided by an embodiment of the present application;

[0076] Figure 2 It is a flowchart of a data processing method provided by an embodiment of the present application;

[0077] Figure 3 It is a flowchart of another data processing method provided by an embodiment of the present application;

[0078] Figure 4a It is a schematic diagram of a data interaction process provided by an embodiment of the present application;

[0079] Figure 4b It is a schematic diagram of another data interaction process provided by an embodiment of the present application;

[0080] Figure 5 It is a flowchart of yet another data processing method provided by an embodiment of the present application;

[0081] Figure 6 It is a flowchart of still another data processing method provided by an embodiment of the present application;

[0082] Figure 7Schematic diagram of a data processing device provided by an embodiment of the present application;

[0083] Figure 8 Schematic diagram of another data processing device provided by an embodiment of the present application;

[0084] Figure 9 Schematic diagram of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0085] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application.

[0086] This application relates to technologies related to artificial intelligence. The related technologies involved are briefly introduced below:

[0087] Artificial Intelligence (AI): The so-called AI is to use a digital computer or a machine controlled by a digital computer to simulate, extend, and expand human intelligence, a theory, method, technology, and application system that can perceive the environment, acquire knowledge, and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. Artificial intelligence also studies the design principles and implementation methods of various intelligent machines, enabling the machines to have the functions of perception, reasoning, and decision-making. The embodiments of the present application mainly involve predicting target network parameters through a risk prediction model to obtain second prediction data, and the second prediction data is used to determine the service parameters of the target service.

[0088] AI technology is an interdisciplinary subject, involving a wide range of fields, including both hardware-level technologies and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, pre-trained model technology, operation / interaction systems, mechatronics, etc. Among them, the pre-trained model is also called the large model or the basic model, and can be widely applied to downstream tasks in various major directions of artificial intelligence after fine-tuning. The artificial intelligence software technology mainly includes several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0089] Machine Learning (ML) is an interdisciplinary field that involves multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve their own performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent, and its applications cover all fields of artificial intelligence. Machine learning and deep learning usually include technologies such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and rote learning. Pre-trained models are the latest development results of deep learning, which integrate the above technologies. The embodiments of this application mainly involve training the model to be trained of a network element (such as an application function network element) based on the total gradient and total loss information to further improve the prediction accuracy of the trained model (such as a risk prediction model).

[0090] Based on the above technologies related to artificial intelligence, the embodiments of this application provide a data processing solution that can jointly train the models of each network element without leaking the data of each network element. Figure 1 A schematic diagram of a network architecture provided for the embodiments of this application, such as Figure 1As shown in the figure, the network architecture provided by this application includes a terminal device 101, an Application Function (AF) network element 102, a Data Warehouse Data Access Function (DWDAF) network element 103, and a Radio Access Network (RAN) network element 104. The data processing solution provided by this application can be executed by the network data analysis network element 103 or the application function network element 102. Among them, the application function network element 102 can specifically be a server, and the radio access network network element 104 can specifically be a base station, a Road Side Unit (RSU), a Wireless Local Area Network (Wi-Fi) device, etc. The network data analysis network element 103 is used to collect data of each network element in the network and analyze and aggregate the collected data. The terminal device can include, but is not limited to: smart phones (such as Android phones, IOS phones, etc.), tablet computers, portable personal computers, Mobile Internet Devices (MIDs), intelligent voice interaction devices, smart home appliances, vehicle-mounted terminals, aircraft, wearable devices, etc. This application embodiment does not make any limitations in this regard; the servers involved can be independent physical servers, or server clusters or distributed systems composed of multiple physical servers, or cloud servers that provide basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms. This application embodiment does not make any limitations in this regard.

[0091] It should be noted that Figure 1 the quantities of the terminal device 101, the application function network element 102, the network data analysis network element 103, and the radio access network network element 104 are only for illustration and do not constitute an actual limitation of this application. The application function network element 102 can be included in the core network or independent of the core network. This application does not make any restrictions in this regard. The terminal device 101, the application function network element 102, and the network data analysis network element 103 can be connected by wired or wireless means. When the terminal device 101 and the application function network element 102 are connected by wired means, the radio access network network element 104 may not be included in the network architecture. This application does not make any restrictions in this regard.

[0092] The general process of the data processing solution provided by this application is as follows:

[0093] (1) The network data analysis network element 103 creates a key pair and sends the first public key to N network elements in the network (such as the terminal device 101, the application function network element 102, and the radio access network network element 104). The first public key is the public key in the key pair, and N is an integer greater than 1.

[0094] (2) The application function network element 102 obtains the first public key and the target sample parameters, and predicts the target sample parameters through the model to be trained to obtain the first prediction data. The target sample parameters are the sample parameters obtained by aligning the sample parameters of the application function network element with the sample parameters of N - 1 network elements (excluding the application function network element) in the network. For example, assume N = 3, the network includes a terminal device, an application function network element, and a radio access network network element. The terminal device includes the sample parameters of users 1 - 5 in the user equipment (UE), the application function network element includes the sample parameters of users 2 - 9 in the application function network element, and the radio access network network element includes the sample parameters of users 1 - 10 in the radio access network network element. Then, the terminal device, the application function network element, and the radio access network network element are aligned to obtain the target sample parameters (in each network element) as the sample parameters of users 2 - 5.

[0095] (3) The application function network element 102 generates the first encrypted information based on the first prediction data and the first public key, and returns the first encrypted information to the network data analysis network element 103. Among them, the first encrypted information is the encrypted information of the gradient and loss information of the application function network element, and the gradient and loss information of the application function network element are calculated based on the first prediction data. In one implementation, the application function network element 102 obtains N - 1 intermediate data sent by N - 1 network elements. The i-th intermediate data is the prediction result obtained by the i-th network element using the model to be trained in the network element for the target sample parameters of the network element; or it is the aggregation result of the target sample parameters of the i-th network element (such as calculating the calculation result for the target sample parameters). i is a positive integer less than or equal to N. The application function network element 102 calculates the gradient and loss information based on the first prediction data and N - 1 intermediate data, and encrypts the gradient and loss information using the first public key to obtain the first encrypted information.

[0096] (4) The network data analysis network element 103 obtains N first encryption messages, generates parameter correction information based on the first private key and the N first encryption messages, and sends the parameter correction information to each network element; wherein, the parameter correction information includes the total gradient and total loss information of the N network elements, and the first private key is the private key in the key pair. In one implementation, the network data analysis network element 103 decrypts the N first encryption messages through the first private key to obtain the gradient and loss information of the N network elements, sums up the gradients of the N network elements, and sums up the loss information of the N network elements to obtain the total gradient and total loss information of the N network elements, and packages the total gradient and total loss information of the N network elements to obtain the parameter correction information.

[0097] (5) The application function network element 102 obtains the parameter correction information sent by the network data analysis network element 103, and updates the parameters of the model to be trained based on the parameter correction information to obtain a risk prediction model.

[0098] In the embodiments of the present application, the network data analysis network element creates a key pair and sends the first public key to N network elements in the network. The application function network element obtains the first public key and target sample parameters, and predicts the target sample parameters through the model to be trained to obtain first prediction data. Based on the first prediction data and the first public key, the first encryption message is generated and returned. The network data analysis network element obtains N first encryption messages, generates parameter correction information based on the first private key and the N first encryption messages. The parameter correction information includes the total gradient and total loss information of the N network elements, and sends the parameter correction information to the application function network element to enable the application function network element to update the parameters of the model to be trained based on the parameter correction information to obtain a risk prediction model. It can be seen that by collecting the gradient and loss information of each network element by the network data analysis network element and returning the total gradient and total loss information obtained based on the gradient and loss information of each network element to each network element, the model of each network element can be jointly trained without disclosing the data (network parameters) of each network element, thereby improving the prediction accuracy of the model of each network element.

[0099] Based on the above data processing solution, the embodiments of the present application propose a more detailed data processing method. The data processing method proposed in the embodiments of the present application will be introduced in detail below with reference to the accompanying drawings.

[0100] Please refer to Figure 2 , Figure 2 which is a flowchart of a data processing method provided by the embodiments of the present application. This data processing method can be executed by a computer device; specifically, the computer device can be the network data analysis network element 103 shown in Figure 1 . As shown in Figure 2 , this data processing method may include but is not limited to S201-S204:

[0101] S201. Create a key pair and send the first public key to N network elements in the network.

[0102] The key pair includes a first public key and a first private key. The first public key is used to encrypt data, and the first private key is used to decrypt the data encrypted by the first public key. Encrypting data with the first public key can ensure that only the holder of the first private key can decrypt the data. N is an integer greater than 1. In one implementation, the N network elements in the network include: user equipment, radio access network elements, and application function network elements.

[0103] S202. Obtain N first encrypted messages.

[0104] The i-th first encrypted message is obtained by the i-th network element encrypting the gradient and loss information of the i-th network element with the first public key. The gradient and loss information of the i-th network element is calculated based on the first prediction data of the network element. The first prediction data is obtained by the training model corresponding to the i-th network element predicting the target sample parameters of the i-th network element. The target sample parameters are the sample parameters obtained after aligning (encrypting) the sample parameters of the i-th network element with the sample parameters of N - 1 network elements. i is a positive integer less than or equal to N. It should be noted that during the alignment process of the sample parameters of each network element, no network element can obtain the plaintext of the sample parameters of other network elements outside this network element.

[0105] In one implementation, the training models corresponding to each network element are deployed locally on the network element, and the computer device obtains N first encrypted messages from the N network elements. In another implementation, the network further includes various Over-The-Top (OTT) servers based on the open Internet, and the training models corresponding to each network element are deployed in the OTT servers. The computer device obtains N first encrypted messages from the OTT servers. In yet another implementation, the network further includes OTT servers. The training models corresponding to some network elements are deployed locally on the network element, and the training models corresponding to other network elements are deployed in the OTT servers. The computer device obtains N first encrypted messages from the network elements with the training models deployed locally and the OTT servers.

[0106] S203. Generate parameter correction information based on the first private key and N first encrypted messages.

[0107] The first private key is the private key in the key pair. The first private key is used to decrypt N first encrypted messages, and the parameter calibration information includes the total gradient and total loss information of N network elements. In one implementation, the computer device decrypts N first encrypted messages with the first private key to obtain the gradient and loss information of N network elements. On the one hand, the computer device sums up the gradients of N network elements to obtain the total gradient of N network elements; on the other hand, the computer device sums up the loss information of N network elements to obtain the total loss information of N network elements. After obtaining the total gradient and total loss information of N network elements, the computer device packages the total gradient and total loss information of N network elements to obtain the parameter calibration information.

[0108] S204. Send the parameter calibration information to each network element.

[0109] In one implementation, the to-be-trained model corresponding to each network element is deployed locally on the network element. The network element includes an application function network element. The computer device sends the parameter calibration information to the application function network element so that the application function network element updates the parameters of the to-be-trained model of the application function network element based on the parameter calibration information.

[0110] In another implementation, the network further includes various video and data (Over-The-Top, OTT) servers based on the open Internet. The to-be-trained model corresponding to each network element is deployed in the OTT server. The computer device sends the parameter calibration information to the OTT server so that the OTT server updates the parameters of the to-be-trained model corresponding to each network element based on the parameter calibration information.

[0111] In yet another implementation, the to-be-trained model corresponding to a part of the network elements is deployed locally on the network element, and the to-be-trained model corresponding to another part of the network elements is deployed in the OTT server. The computer device sends the parameter calibration information to the network elements with the to-be-trained model deployed locally and the OTT server so that the network elements with the to-be-trained model deployed locally and the OTT server update the parameters of the to-be-trained model corresponding to each network element based on the parameter calibration information.

[0112] In the embodiments of the present application, a key pair is created, and the first public key is sent to N network elements in the network, and N first encrypted messages are obtained. The i-th first encrypted message is obtained by encrypting the gradient and loss information of the i-th network element. Based on the first private key and the N first encrypted messages, parameter correction information is generated. The parameter correction information includes the total gradient and total loss information of the N network elements. The parameter correction information is sent to each network element so that each network element updates the parameters of the model to be trained of the network element based on the parameter correction information. It can be seen that by collecting the gradient and loss information of each network element through a third party (such as a network data analysis network element) and returning the obtained total gradient and total loss information to each network element, the models of each network element can be jointly trained without disclosing the data (network parameters) of each network element, thereby improving the prediction accuracy of the models of each network element.

[0113] Please refer to Figure 3 , Figure 3 which is a flowchart of another data processing method provided by the embodiments of the present application. This data processing method can be executed by a computer device; specifically, the computer device can be Figure 1 the network data analysis network element 103 shown in Figure 3 . As shown in

[0114] S301. Obtain a service subscription request.

[0115] The service subscription request is used to request collaborative training of the models to be trained of N network elements, where N is an integer greater than 1. In one implementation, the service subscription request carries the identifiers of the N network elements participating in the collaborative training and the identity information of the sender of the service subscription request.

[0116] S302. If the identity information passes the verification, return response information to the service subscription request.

[0117] In one implementation, after receiving the service subscription request, the computer device verifies the identity information of the sender of the service subscription request. The verification content may include, but is not limited to: whether the sender of the service subscription request has the permission to subscribe to the service (such as whether the identity information of the sender of the subscription request exists in the permission list). If the identity information of the sender of the service subscription request passes the verification, the computer device returns response information to the sender of the service subscription request. The response information is used to indicate that the service subscription request has passed; correspondingly, if the identity information of the sender of the service subscription request fails to pass the verification, the computer device returns rejection information to the sender of the service subscription request. The rejection information is used to indicate that the service subscription request has not passed.

[0118] In one embodiment, if the identity information of the sender of the service subscription request is verified, the computer device may further send confirmation information to N - 1 network elements other than the sender of the service subscription request based on the identifiers of the N network elements participating in the collaborative training to confirm whether the N - 1 network elements agree to participate in the collaborative training.

[0119] S303. Create a key pair and send the first public key to N network elements in the network.

[0120] S304. Obtain N first encrypted messages.

[0121] S305. Generate parameter correction information based on the first private key and the N first encrypted messages.

[0122] S306. Send the parameter correction information to each network element.

[0123] For the specific implementation manners of S303 - S306, reference may be made to Figure 2 the implementation methods of S201 - S204 in , which will not be elaborated here.

[0124] S307. Obtain N second encrypted messages.

[0125] The i-th second encrypted message is obtained by the i-th network element encrypting the second prediction data of the network element using the first public key. The second prediction data is obtained by predicting the target network parameters of the i-th network element using the i-th sub-model (the collaboratively trained model corresponding to the i-th network element). The target network parameters are the network parameters obtained by aligning the network parameters of the i-th network element with the network parameters of N - 1 network elements.

[0126] In one implementation manner, the computer device obtains a service parameter request. The service parameter request is used to request the acquisition of service parameters associated with a target service. In one embodiment, the service parameter request carries the identity information of the sender of the service parameter request. After receiving the service parameter request, the computer device verifies the identity information of the sender of the service parameter request. The verification content may include, but is not limited to: whether the sender of the service parameter request has the permission to acquire the service parameters. If the identity information of the sender of the service parameter request is verified, on the one hand, the computer device returns a response message to the sender of the service parameter request, and the response message is used to indicate that the service parameter request has passed. On the other hand, the computer device obtains the N second encrypted messages required to determine the service parameters associated with the target service and continues to execute S308. Correspondingly, if the identity information of the sender of the service parameter request fails to pass the verification, the computer device returns a prompt message to the sender of the service parameter request, and the prompt message is used to indicate that the service parameter request has not passed.

[0127] In another embodiment, the N network elements include a user equipment, a radio access network element, and an application function network element. The second encryption information of the user equipment is obtained by the computer device from the user equipment; similarly, the second encryption information of the radio access network element is obtained by the computer device from the radio access network element, and the second encryption information of the application function network element is obtained by the computer device from the application function network element.

[0128] In another embodiment, the N network elements include a user equipment, a radio access network element, and an application function network element, and the network further includes various video and data (Over-The-Top, OTT) servers based on the open Internet. The computer device obtains the second encryption information of the user equipment, the second encryption information of the radio access network element, and the second encryption information of the application function network element from the OTT server.

[0129] In one embodiment, the second encryption information of any network element is obtained by the OTT server encrypting the second prediction data of the network element using the first public key. The second prediction data is obtained by the OTT server predicting the target network parameters of the network element using the co-trained model (corresponding to the network element). The target network parameters of any network element are provided by the network element to the OTT server. That is to say, the models corresponding to each network element can be deployed in the OTT server. It should be noted that the network element can use the homomorphic encryption method to provide encrypted data to the OTT server to ensure the privacy of its own network parameters. Taking the network element as the user equipment as an example, the second encryption information of the user equipment is obtained by the OTT server encrypting the second prediction data of the user equipment using the first public key. The second prediction data is obtained by the OTT server predicting the target network parameters of the user equipment using the co-trained model (corresponding to the user equipment). The target network parameters of the user equipment are provided by the user equipment.

[0130] In another embodiment, each OTT server is associated with a region, and the second encryption information of any network element is collected by the OTT server associated with the region to which the network element belongs from the network element. That is to say, the models corresponding to each network element can be deployed locally. Taking the network element as the radio access network element as an example, the second encryption information of the radio access network element is obtained by the radio access network element encrypting the second prediction data of the network element using the first public key. The second prediction data is obtained by the radio access network element predicting the target network parameters of the radio access network element using the local sub-model (co-trained model). The target network parameters of the radio access network element are the network parameters obtained by aligning the network parameters of the radio access network element with the network parameters of N-1 network elements.

[0131] In yet another embodiment, each OTT server is associated with a region. The second encryption information of a part of network elements is obtained by the OTT server encrypting the second prediction data of the network element using the first public key. The second prediction data is obtained by the OTT server predicting the target network parameters of the network element using the co-trained model (corresponding to the network element). The second encryption information of another part of network elements is collected by the OTT server associated with the region to which the network element belongs. For example, the second encryption information of the user equipment is obtained by the OTT server encrypting the second prediction data of the user equipment using the first public key. The second prediction data is obtained by the OTT service predicting the target network parameters of the user equipment using the co-trained model (corresponding to the user equipment). The second encryption information of the radio access network element and the application function network element is collected by the OTT server from the radio access network element and the application function network element.

[0132] The network parameters of the user equipment include at least one of the following: per user Signal to Interference plus Noise Ratio (perUE-SINR), per user Received Signal Strength Indicator (perUE-RSSI), per user Reference Signal Receiving Power (perUE-RSRP), per user Reference Signal Receiving Quality (perUE-RSRQ), per user latency, per user data rate. The network parameters of the radio access network network element include at least one of the following: Signal to Interference and Noise Ratio (SINR), Received Signal Strength Indicator (RSSI), Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), latency, data rate, base station (cell) identifier (cell-id). The network parameters of the application function network element are determined based on the target service; for example, if the target service is an audio-video service, the network parameters of the application function network element may include at least one of the following: encoding and decoding information, video bit rate and frame rate information, media transmission control information, media synchronization information; for another example, if the target service is a vehicle networking service, the network parameters of the application function network element may include at least one of the following: vehicle status information, driving behavior information, road condition information, environmental information, location information, driver information.

[0133] Optionally, the computer device obtaining the N second encryption information may also refer to: the computer device obtaining the encryption result obtained by aggregating the N second encryption information (that is, not obtaining the second encryption information of a single network element).

[0134] S308. Determine the service parameters associated with the target service based on the first private key and the N second encryption information.

[0135] In one implementation, the computer device decrypts N pieces of second encrypted information with the first private key to obtain N pieces of plaintext data. In one embodiment, the computer device may directly determine the N pieces of plaintext data as the service parameters associated with the target service. In another embodiment, the computer device performs integration processing on the N pieces of plaintext data to obtain the service parameters associated with the target service, and the integration processing includes at least one of the following: data format conversion, data aggregation, data intersection, data filtering, and data analysis.

[0136] In another implementation, on the one hand, the computer device decrypts N pieces of second encrypted information with the first private key to obtain N pieces of plaintext data; on the other hand, the computer device obtains M pieces of third encrypted information and performs aggregation processing on the M pieces of third encrypted information to obtain an aggregation result corresponding to the M pieces of third encrypted information; wherein, the third encrypted information is obtained by a network element in the network encrypting (homomorphically) the network parameters of the network element with the second public key, and the second public key is provided by the i-th network element (the acquirer of the service parameters associated with the target service), and M is an integer greater than 1.

[0137] After obtaining the N pieces of plaintext data and the aggregation result, the computer device packages the aggregation result and the N pieces of plaintext data to obtain the service parameters associated with the target service. That is to say, a part of the service parameters associated with the target service can be obtained by predicting the target network parameters of the network element using a model, and another part can be obtained by aggregating the target network parameters of multiple network elements.

[0138] It should be noted that the aggregation parameters obtained after decrypting the aggregation result match the target result, and the target result is obtained by aggregating the M pieces of plaintext data corresponding to the M pieces of third encrypted information. As can be seen from the above, the N pieces of plaintext data are obtained by predicting the network parameters of the network element using a model, and the network parameters of the network element are not exposed. The aggregation result is obtained by aggregating the network parameters of multiple network elements. After the acquirer of the aggregation result decrypts the aggregation result, the obtained aggregation parameters are obtained by aggregating the network parameters of multiple network elements, and the network parameters of a single network element are not exposed.

[0139] Optionally, the computer device obtaining N pieces of second encrypted information means that the computer device obtains the encrypted result obtained after aggregating the N pieces of second encrypted information; in this case, the computer device decrypts the encrypted result obtained after aggregating the N pieces of second encrypted information with the first private key to obtain the result of aggregating the N pieces of second prediction data.

[0140] S309. Send service parameters to the i-th network element.

[0141] In one implementation, the i-th network element is an application function network element, and the computer device sends service parameters to the application function network element so that the application function network element adjusts the service policy of the target service based on the service parameters.

[0142] Figure 4a This is a schematic diagram of a data interaction process provided by an embodiment of the present application. As Figure 4a shown, the network includes an application function network element, a network data analysis network element, a radio access network element, and a user equipment. The application function network element can trigger the joint training of the to-be-trained model by sending a service subscription request to the network data analysis network element. The training process includes: the network data analysis network element obtains the first encrypted data sent by the application function network element, the radio access network element, and the user equipment. The first encrypted data of any network element is obtained by encrypting the first prediction data with the public key provided by the network data analysis network element. The first prediction data is obtained by the network element calling the corresponding to-be-trained model to predict the target sample parameters of the network element. After obtaining N pieces of first encrypted data, the network data analysis network element generates parameter correction information based on the N pieces of first encrypted data. The specific implementation manner can refer to S203 and will not be elaborated here. After generating the parameter correction information, the network data analysis network element can return the parameter correction information to each network element so that each network element trains the to-be-trained model (local) based on the parameter correction information (such as adjusting the parameters of the to-be-trained model).

[0143] Further, the application function network element can obtain service parameters by sending a service parameter request to the network data analysis network element. The service parameter obtaining process includes: the network data analysis network element obtains the second encrypted data sent by the application function network element, the radio access network element, and the user equipment. The second encrypted data of any network element is obtained by encrypting the second prediction data with the public key provided by the network data analysis network element. The second prediction data is obtained by the network element calling the corresponding trained model to predict the target network parameters of the network element. After obtaining N pieces of second encrypted data, the network data analysis network element determines the service parameters based on the N pieces of second encrypted data. The specific implementation manner can refer to S308 and will not be elaborated here. After determining the service parameters, the network data analysis network element can return the service parameters to the application function network element, and the application function network element can adjust the service policy based on the returned service parameters (such as sending a service adjustment instruction to the user equipment).

[0144] Figure 4b This is another schematic diagram of a data interaction process provided by an embodiment of the present application. As Figure 4b shown, the network includes an application function network element, a Network Exposure Function (NEF), a network data analysis network element, a radio access network element, an OTT server, and a user equipment. Compared withFigure 4a For the NEF, it is used to forward data between the application function network element and the network data analysis network element. The OTT server can be used to collect (or generate) the first encrypted data / second encrypted data of one or more network elements. The to-be-trained model (trained model) corresponding to each network element can be deployed locally or in the OTT server. For the specific data interaction process, reference can be made to Figure 4a , which will not be elaborated here.

[0145] In the embodiments of the present application, a key pair is created, and the first public key is sent to N network elements in the network to obtain N first encrypted messages. The i-th first encrypted message is obtained by encrypting the gradient and loss information of the i-th network element. Based on the first private key and the N first encrypted messages, parameter correction information is generated. The parameter correction information includes the total gradient and total loss information of the N network elements. The parameter correction information is sent to each network element so that each network element updates the parameters of the to-be-trained model of the network element based on the parameter correction information. It can be seen that by collecting the gradient and loss information of each network element through a third party (such as the network data analysis network element) and returning the obtained total gradient and total loss information to each network element, the joint training of the models of each network element can be performed without disclosing the data (network parameters) of each network element, thereby improving the prediction accuracy of the models of each network element. In addition, by obtaining N second encrypted messages and determining the service parameters associated with the target service based on the first private key and the N second encrypted messages, the data of each network element can be aggregated without disclosing the data (network parameters) of each network element, thereby reducing the service decision-making errors caused by the incomplete network data of a single network element.

[0146] Please refer to Figure 5 , Figure 5 which is a flowchart of another data processing method provided by the embodiments of the present application. This data processing method can be executed by a computer device; the computer device can specifically be Figure 1 the application function network element 102 shown in Figure 5 , or it can also be the OTT server. As

[0147] S501. Obtain the first public key and the target sample parameters.

[0148] The network includes N network elements, where N is an integer greater than 1. The target sample parameter is the sample parameter obtained by aligning the sample parameter of the application function network element with the sample parameters of N - 1 network elements in the network. For example, let N = 3. The network includes a terminal device, an application function network element, and a radio access network element. The terminal device includes the sample parameters of users 1 - 5 in the user equipment (UE). The application function network element includes the sample parameters of users 2 - 9 in the application function network element. The radio access network element includes the sample parameters of users 1 - 10 in the radio access network element. Then, by aligning the terminal device, the application function network element, and the radio access network element, the obtained target sample parameter (in each network element) is the sample parameter of users 2 - 5. The first public key can be sent by the network data analysis network element to the computer device after creating the key pair.

[0149] S502. Use the model to be trained to predict the target sample parameter to obtain the first prediction data.

[0150] In one implementation, the model to be trained is deployed locally on the computer. The computer device calls the model to be trained to predict the target sample parameter to obtain the first prediction data.

[0151] In another implementation, the network further includes an OTT server. The model to be trained corresponding to the application function network element is deployed in the OTT server. The computer device sends the (homomorphically) encrypted target sample parameter to the OTT server, so that the OTT server calls the model to be trained corresponding to the application function network element to predict the (homomorphically) encrypted target sample parameter to obtain the encrypted prediction result. The computer device obtains the encrypted prediction result returned by the OTT server and decrypts the encrypted prediction result to obtain the first prediction data. It can be seen that a network element (such as an application function network element) can use the homomorphic encryption method to provide the homomorphically encrypted data (such as the target sample parameter) to the OTT server, so as to call the model to be trained in the OTT server to predict the homomorphically encrypted data (such as the target sample parameter) while ensuring the privacy of the target sample parameter.

[0152] S503. Generate the first encrypted information based on the first prediction data and the first public key, and return the first encrypted information to the provider of the first public key.

[0153] The first encrypted information is obtained by encrypting the gradient and loss information of the application function network element, and the gradient and loss information of the application function network element are calculated based on the first prediction data.

[0154] In one embodiment, a computer device obtains N - 1 intermediate data sent by N - 1 network elements. The i-th intermediate data is obtained by the i-th network element predicting the target sample parameters of the network element using the model to be trained corresponding to the network element; or is obtained by the i-th network element aggregating the target sample parameters of the network element (such as calculating a calculation result from the target sample parameters), where i is a positive integer less than or equal to N. The computer device calculates gradients and loss information based on the first prediction data and the N - 1 intermediate data, and encrypts the gradients and loss information using the first public key to obtain the first encrypted information.

[0155] S504. Obtain parameter correction information sent by the provider of the first public key.

[0156] The parameter correction information includes the total gradients and total loss information of N network elements. The total gradients and total loss information of the N network elements are calculated based on the gradients and loss information of the application function network elements.

[0157] S505. Update the parameters of the model to be trained based on the parameter correction information to obtain a risk prediction model.

[0158] In one embodiment, the computer device adjusts the parameters in the model to be trained based on the parameter correction information, and obtains updated parameter correction information after the parameter adjustment according to the implementation manner in S502 - S504, and iterates repeatedly until the loss value indicated by the total loss information is less than the loss threshold, to obtain a risk prediction model (i.e., the trained model).

[0159] In the embodiments of the present application, the first public key and the target sample parameters are obtained, the target sample parameters are predicted through the model to be trained to obtain the first prediction data, the first encrypted information is generated based on the first prediction data and the first public key, and the first encrypted information is returned to the provider of the first public key. The first encrypted information is obtained by encrypting the gradients and loss information of the application function network elements. The parameter correction information sent by the provider of the first public key is obtained. The parameter correction information includes the total gradients and total loss information of N network elements. The parameters of the model to be trained are updated based on the parameter correction information to obtain a risk prediction model. It can be seen that by providing the gradients and loss information to a third party (such as a network data analysis network element), the application function network element can obtain the total gradients and total loss information returned by the third party based on the gradients and loss information of each network element, and update the parameters of the model to be trained based on the total gradients and total loss information, so as to achieve the purpose of jointly training the models of each network element without disclosing the data of each network element.

[0160] Please refer to Figure 6 , Figure 6 which is a flowchart of another data processing method provided by the embodiments of the present application. This data processing method can be executed by a computer device; specifically, the computer device can beFigure 1 The application function network element 102 shown in the figure. As Figure 6 shown, the data processing method may include but is not limited to S601 - S610:

[0161] S601. Send a service subscription request to the network data analysis network element.

[0162] When the computer device needs to train the model to be trained, the network data analysis network element sends a service subscription request. The service subscription request is used to request collaborative training of the models to be trained of N network elements, where N is an integer greater than 1. In one implementation, the service subscription request carries the identifiers of the N network elements participating in the collaborative training and the identity information of the sender of the service subscription request. Correspondingly, the computer device can obtain the response information (or rejection information) returned by the network data analysis network element. The response information is used to indicate that the service subscription request is passed (the rejection information is used to indicate that the service subscription request is not passed).

[0163] S602. Obtain the first public key and the target sample parameters.

[0164] S603. Use the model to be trained to predict the target sample parameters to obtain the first prediction data.

[0165] S604. Generate the first encrypted information based on the first prediction data and the first public key, and return the first encrypted information to the provider of the first public key.

[0166] S605. Obtain the parameter correction information sent by the provider of the first public key.

[0167] S606. Update the parameters of the model to be trained based on the parameter correction information to obtain the risk prediction model.

[0168] For the specific implementation manners of S602 - S606, reference may be made to Figure 5 the implementation manners of S501 - S505 in, which will not be elaborated here.

[0169] S607. Use the risk prediction model to predict the target network parameters to obtain the second prediction data.

[0170] In one implementation, the computer device obtains the target network parameters and uses the risk prediction model (i.e., the trained model) to predict the target network parameters to obtain the second prediction data; where the target network parameters are the network parameters obtained by aligning the network parameters of the application function network element with the network parameters of N - 1 network elements in the network.

[0171] S608. Encrypt the second prediction data with the first public key to obtain the second encrypted information, and return the second encrypted information to the provider of the first public key.

[0172] In one implementation, after obtaining the second encrypted information, the computer device directly sends the second encrypted information to the network data analysis network element.

[0173] In another implementation, the computer device sends the second encrypted information to a target network element in the network, so that the target network element aggregates its own second encrypted information and the second encrypted information provided by the computer device, and sends the aggregated encrypted result to the network data analysis network element.

[0174] In yet another implementation, the computer device obtains the second encrypted information sent by N - 1 network elements, aggregates the N second encrypted information (including the N - 1 second encrypted information obtained and the second encrypted information obtained by encrypting the second prediction data with the first public key), and sends the aggregated encrypted result to the network data analysis network element.

[0175] It can be understood that, compared with directly sending the second encrypted information to the network data analysis network element, sending the aggregated encrypted result to the network data analysis network element can prevent the leakage of the second prediction data of a single network element and further improve data privacy.

[0176] S609: Obtain service parameters sent by the provider of the first public key.

[0177] The service parameters are associated with the target service. For example, when the target service is a network quality - related service such as remote tele - driving or multimedia data transmission, the service parameters may include the network quality at a certain future moment; when the target service is a network security - related service such as information management, the service parameters may include the probability of a network storm occurring at a certain future moment. The subject of a network storm can spread quickly through channels such as the network and social media, and trigger network attention and discussions; specifically, the subject of a network storm can be a large - scale event that spreads rapidly on the network, which may involve false information, malware, cyber - attacks, etc., or can be used to describe a hot topic or event on the network (such as a popular topic on social media or a viral video, etc.).

[0178] S610: Adjust the service strategy of the target service based on the service parameters.

[0179] In one implementation, the target service is a remote teleoperation service, and the service parameters include the communication quality at the first moment. Remote teleoperation operates a controlled object (such as a vehicle, an aircraft, a robot, etc.) through a remote control device (such as a remote controller, a computer, etc.). This technology is usually applied in fields such as driverless vehicles, robotic vehicles, remote control toy cars, etc., and can operate the controlled object in extreme situations (such as in a dangerous environment), improving driving safety and efficiency. If the communication quality at the first moment is lower than the quality threshold, the computer device reduces the driving speed of the remotely controlled object at the second moment, where the second moment is before the first moment (i.e., decelerate in advance when the communication quality is poor). In addition, safety measures such as adjusting the priority of autonomous driving and manual control can be taken to ensure driving safety when the communication quality is poor. This application does not limit the safety measures taken.

[0180] In another implementation, the target service is information management, and the service parameters include the probability of a network storm occurring at the first moment. If the probability of a network storm occurring at the first moment is greater than the probability threshold, the computer device takes a restriction measure at the second moment (such as restricting the propagation of the network storm entity in the network), where the second moment is before the first moment (i.e., take a restriction measure on the network storm entity in advance before the network storm occurs).

[0181] In yet another implementation, the service parameters include an aggregation result and N plaintext data. The N plaintext data are obtained by decrypting the first encryption information of N network elements, the aggregation result is obtained by aggregating M third encryption information, and the j-th third encryption information is obtained by the j-th network element encrypting the network parameters of the network element using the second public key provided by the computer device (application function network element), where j is a positive integer less than or equal to M. After obtaining the service parameters, the computer device decrypts the aggregation result using the second private key corresponding to the second public key to obtain aggregation parameters, and adjusts the service strategy of the target service based on the aggregation parameters and the N plaintext data.

[0182] In the embodiments of the present application, a first public key and target sample parameters are obtained, the target sample parameters are predicted by a model to be trained to obtain first prediction data, first encrypted information is generated based on the first prediction data and the first public key, and the first encrypted information is returned to the provider of the first public key. The first encrypted information is obtained by encrypting the gradient and loss information of the application function network element. The parameter correction information sent by the provider of the first public key is obtained. The parameter correction information includes the total gradient and total loss information of N network elements. The parameters of the model to be trained are updated based on the parameter correction information to obtain a risk prediction model. It can be seen that by providing the gradient and loss information to a third party (such as a network data analysis network element), the application function network element can obtain the total gradient and total loss information returned by the third party based on the gradient and loss information of each network element, and update the parameters of the model to be trained based on the total gradient and total loss information, so as to achieve the purpose of jointly training the models of each network element without disclosing the data of each network element. In addition, by sending second encrypted data and adjusting the service strategy of the target service based on the service parameters (obtained from the second encrypted data), the data of each network element can be aggregated without disclosing the data (network parameters) of each network element, thereby reducing the business decision-making errors caused by incomplete network data of a single network element.

[0183] The method of the embodiments of the present application is described in detail above. To facilitate better implementation of the above solutions of the embodiments of the present application, correspondingly, the device of the embodiments of the present application is provided below.

[0184] Please refer to Figure 7 , Figure 7 which is a schematic structural diagram of a data processing device provided by the embodiments of the present application. The device can be mounted on a computer device, and the computer device can specifically be Figure 1 the network data analysis network element 103 shown in the figure. Figure 7 The data processing device shown in the figure can be used to execute some or all of the functions described in the above Figure 2 and Figure 3 method embodiments. Please refer to Figure 7 , and the detailed descriptions of each unit are as follows:

[0185] A processing unit 701 is configured to create a key pair and send a first public key to N network elements in the network through a sending unit 702. The first public key is the public key in the key pair, and N is an integer greater than 1;

[0186] An obtaining unit 703, configured to obtain N first encrypted messages. The i-th first encrypted message is obtained by encrypting the gradient and loss information of the i-th network element using a first public key. The gradient and loss information of the i-th network element is calculated based on the first prediction data of the network element. The first prediction data is obtained by predicting the target sample parameters of the i-th network element using a model to be trained. The target sample parameters are the sample parameters obtained by aligning the sample parameters of the i-th network element with the sample parameters of N - 1 network elements. i is a positive integer less than or equal to N;

[0187] A processing unit 701 is further configured to generate parameter correction information based on the first private key and N first encrypted messages. The parameter correction information includes the total gradient and total loss information of N network elements. The first private key is the private key in the key pair;

[0188] A sending unit 702 is further configured to send the parameter correction information to each network element, so that each network element updates the parameters of the model to be trained of the network element based on the parameter correction information.

[0189] In one implementation, the processing unit 701 is further configured to:

[0190] Obtain N second encrypted messages. The i-th second encrypted message is obtained by the i-th network element encrypting the second prediction data of the network element using the first public key. The second prediction data is obtained by predicting the target network parameters of the i-th network element using the i-th sub-model; the target network parameters are the network parameters obtained by aligning the network parameters of the i-th network element with the network parameters of N - 1 network elements;

[0191] Determine the service parameters associated with the target service based on the first private key and N second encrypted messages;

[0192] Send the service parameters to the i-th network element, so that the i-th network element adjusts the service strategy of the target service based on the service parameters.

[0193] In one implementation, the processing unit 701 is configured to determine the service parameters associated with the target service based on the first private key and N second encrypted messages. Specifically, it is configured to:

[0194] Decrypt the N second encrypted messages using the first private key to obtain N plaintext data;

[0195] Obtain M third encrypted messages. The third encrypted message is obtained by a network element in the network encrypting the network parameters of the network element using a second public key. The second public key is provided by the i-th network element. M is an integer greater than 1;

[0196] Aggregate the M third encrypted messages to obtain an aggregation result corresponding to the M third encrypted messages;

[0197] Pack the aggregation result and N plaintext data to obtain service parameters related to the target service.

[0198] In one implementation, the processing unit 701 is configured to determine service parameters related to the target service based on the first private key and N second encrypted messages. Specifically, it is configured to:

[0199] Decrypt the N second encrypted messages with the first private key to obtain N plaintext data;

[0200] Perform integration processing on the N plaintext data to obtain service parameters related to the target service. The integration processing includes at least one of the following: data format conversion, data aggregation, data intersection, data filtering, and data analysis.

[0201] In one implementation, the processing unit 701 is configured to generate parameter correction information based on the first private key and N first encrypted messages. Specifically, it is configured to:

[0202] Decrypt the N first encrypted messages with the first private key to obtain the gradients and loss information of N network elements;

[0203] Sum the gradients of the N network elements to obtain the total gradient of the N network elements;

[0204] Sum the loss information of the N network elements to obtain the total loss information of the N network elements;

[0205] Pack the total gradient and total loss information of the N network elements to obtain parameter correction information.

[0206] In one implementation, the processing unit 701 is further configured to:

[0207] Obtain a service subscription request, which carries the identifiers of N network elements participating in collaborative training and the identity information of the sender of the service subscription request;

[0208] Verify the identity information;

[0209] If the identity information passes the verification, return response information to the sender of the service subscription request. The response information is used to indicate that the service subscription request has passed.

[0210] In one implementation, the network elements in the network include user equipment, radio access network elements, and application function network elements;

[0211] The network parameters of the user equipment include at least one of the following: single-user signal-to-interference-plus-noise ratio, single-user received signal strength indication, single-user reference signal received power, single-user reference signal received quality, single-user delay, and single-user data rate;

[0212] The network parameters of the radio access network element include at least one of the following: signal-to-interference-plus-noise ratio, received signal strength indication, reference signal received power, reference signal received quality, delay, data rate, base station identifier;

[0213] The network parameters of the application function network element are determined based on the target service.

[0214] According to an embodiment of the present application, Figure 2 and Figure 3 Some of the steps involved in the data processing method shown can be executed by each unit in the Figure 7 data processing device shown. For example, Figure 2 S201 shown in Figure 7 can be jointly executed by the processing unit 701 and the sending unit 702 shown, S202 can be executed by the Figure 7 acquiring unit 703 shown, S203 can be executed by the Figure 7 processing unit 701 shown, and S204 can be executed by the Figure 7 sending unit 702 shown. Figure 3 S301, S304, and S307 shown in Figure 7 can be executed by the Figure 7 acquiring unit 703 shown, S302, S306, and S309 can be executed by the Figure 7 sending unit 702 shown, S305 and S308 can be executed by the Figure 7 processing unit 701 shown, and S303 can be jointly executed by the Figure 7 processing unit 701 and the sending unit 702 shown.

[0215] According to another embodiment of the present application, it is possible to construct, for example, on a general computing device such as a computer including processing elements and storage elements such as a central processing unit (CPU), a random access storage medium (RAM), and a read-only storage medium (ROM), by running a computer program (including program code) that can execute the respective steps involved in the corresponding methods shown in Figure 2 and Figure 3 and Figure 7The data processing device shown, and to implement the data processing method of the embodiments of the present application. The computer program can be recorded on, for example, a computer-readable recording medium, and loaded into the above computing device through the computer-readable recording medium, and run therein.

[0216] Based on the same inventive concept, the principle of solving problems and the beneficial effects of the data processing device provided in the embodiments of the present application are similar to the principle of solving problems and the beneficial effects of the data processing method in the method embodiments of the present application. One can refer to the principle of implementation and the beneficial effects of the method. For the sake of brevity, it will not be elaborated here.

[0217] Please refer to Figure 8 , Figure 8 which is a schematic structural diagram of another data processing device provided in the embodiments of the present application. This device can be mounted on a computer device, and the computer device can specifically be Figure 1 the application function network element 102 shown. Figure 8 The data processing device shown can be used to execute some or all of the functions in the method embodiments described above Figure 5 and Figure 6 . Please refer to Figure 8 , and the detailed description of each unit is as follows:

[0218] An acquisition unit 801, configured to acquire a first public key and target sample parameters. The target sample parameters are sample parameters obtained by aligning the sample parameters of the application function network element with the sample parameters of N - 1 network elements in the network, where N is an integer greater than 1;

[0219] A processing unit 802, configured to predict the target sample parameters through a model to be trained, and obtain first prediction data;

[0220] and configured to generate first encrypted information based on the first prediction data and the first public key, and return the first encrypted information to the provider of the first public key through a sending unit 803. The first encrypted information is obtained by encrypting the gradient and loss information of the application function network element, and the gradient and loss information of the application function network element are calculated based on the first prediction data;

[0221] The acquisition unit 801 is further configured to acquire parameter correction information sent by the provider of the first public key. The parameter correction information includes the total gradient and total loss information of N network elements, and the total gradient and total loss information of the N network elements are calculated based on the gradient and loss information of the application function network element;

[0222] The processing unit 802 is further configured to update the parameters of the model to be trained based on the parameter correction information, and obtain a risk prediction model.

[0223] In an implementation manner, the processing unit 802 is further configured to:

[0224] Obtain target network parameters, where the target network parameters are network parameters obtained by aligning the network parameters of the application function network element with the network parameters of N - 1 network elements in the network;

[0225] Use a risk prediction model to predict the target network parameters to obtain second prediction data;

[0226] Encrypt the second prediction data with the first public key to obtain second encrypted information, and return the second encrypted information to the provider of the first public key;

[0227] Obtain service parameters sent by the provider of the first public key, where the service parameters are associated with the target service;

[0228] Adjust the service strategy of the target service based on the service parameters.

[0229] In one implementation, the service parameters include an aggregation result and N plaintext data. The N plaintext data are obtained by decrypting the first encrypted information of N network elements, and the aggregation result is obtained by aggregating M third encrypted information. The jth third encrypted information is obtained by the jth network element encrypting the network parameters of the network element with the second public key provided by the application function network element, where j is a positive integer less than or equal to M;

[0230] The processing unit 802 is configured to adjust the service strategy of the target service based on the service parameters, specifically:

[0231] Decrypt the aggregation result with the second private key corresponding to the second public key to obtain aggregation parameters;

[0232] Adjust the service strategy of the target service based on the aggregation parameters and the N plaintext data.

[0233] In one implementation, the target service is a remote control driving service, and the service parameters include the communication quality at the first moment; the processing unit 802 is configured to adjust the service strategy of the target service based on the service parameters, specifically:

[0234] If the communication quality at the first moment is lower than the quality threshold, then reduce the driving speed of the object to be remotely controlled at the second moment, where the second moment is before the first moment.

[0235] In one implementation, the processing unit 802 is configured to generate first encrypted information based on the first prediction data and the first public key, specifically:

[0236] Obtain N - 1 intermediate data sent by N - 1 network elements; the ith intermediate data is obtained by the ith network element predicting the target sample parameters of the network element using the model to be trained in the network element, or is obtained by the ith network element aggregating the target sample parameters of the network element;

[0237] Calculate the gradient and loss information based on the first prediction data and N-1 intermediate data;

[0238] Encrypt the gradient and loss information using the first public key to obtain the first encrypted information.

[0239] According to an embodiment of the present application, Figure 5 and Figure 6 Some of the steps involved in the data processing method shown can be executed by each unit in the data processing device shown. For example, Figure 8 S501 and S504 shown in can be executed by the acquisition unit 801 shown in, S502 and S505 can be executed by the processing unit 802 shown in, and S503 can be executed jointly by the processing unit 802 and the sending unit 803 shown in. Figure 5 S501 and S504 shown in Figure 8 the acquisition unit 801 shown in, S502 and S505 can be executed by the processing unit 802 shown in, and S503 can be executed jointly by the processing unit 802 and the sending unit 803 shown in. Figure 8 S502 and S505 can be executed by the processing unit 802 shown in, and S503 can be executed jointly by the processing unit 802 and the sending unit 803 shown in. Figure 8 the processing unit 802 and the sending unit 803 shown in. Figure 6 S602, S605 and S609 shown in Figure 8 can be executed by the acquisition unit 801 shown in, S603, S606, S607 and S610 can be executed by the processing unit 802 shown in, S601 can be executed by the sending unit 803 shown in, and S604 and S608 can be executed jointly by the processing unit 802 and the sending unit 803 shown in. Figure 8 S602, S605 and S609 shown in Figure 8 the acquisition unit 801 shown in, S603, S606, S607 and S610 can be executed by the processing unit 802 shown in, S601 can be executed by the sending unit 803 shown in, and S604 and S608 can be executed jointly by the processing unit 802 and the sending unit 803 shown in. Figure 8 the processing unit 802 and the sending unit 803 shown in. Figure 8 Each unit in the data processing device shown can be separately or all combined into one or several other units to form, or some of the units can be further split into multiple smaller units with functional division to form, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above units are divided based on logical functions. In actual applications, the function of one unit can also be realized by multiple units, or the functions of multiple units can be realized by one unit. In other embodiments of the present application, the data processing device can also include other units. In actual applications, these functions can also be assisted by other units and can be realized by the cooperation of multiple units.

[0240] According to another embodiment of the present application, it can be constructed by running a computer program (including program code) capable of executing the respective steps involved in the corresponding methods shown in and on a general computing device such as a computer including processing elements and storage elements such as a central processing unit (CPU), a random access storage medium (RAM), and a read-only storage medium (ROM). Figure 5 and Figure 6 shown in Figure 8The data processing device shown in the figure, and to implement the data processing method of the embodiments of the present application. The computer program can be recorded on, for example, a computer-readable recording medium, and loaded into the above computing device through the computer-readable recording medium, and run therein.

[0241] Based on the same inventive concept, the principle and beneficial effects of the data processing device provided in the embodiments of the present application for solving problems are similar to those of the data processing method in the method embodiments of the present application. The principle and beneficial effects of the method implementation can be referred to. For the sake of brevity, they will not be repeated here.

[0242] Please refer to Figure 9 , Figure 9 which is a schematic structural diagram of a computer device provided in the embodiments of the present application. As Figure 9 shown, the computer device at least includes a processor 901, a communication interface 902, and a memory 903. Among them, the processor 901, the communication interface 902, and the memory 903 can be connected through a bus or other means. The processor 901 (or Central Processing Unit, CPU) is the computing core and control core of the computer device. It can parse various instructions in the computer device and process various data of the computer device. For example, the CPU can be used to parse the power-on and power-off instructions sent by the user to the computer device and control the computer device to perform power-on and power-off operations. Another example is that the CPU can transmit various interactive data between the internal structures of the computer device, and so on. The communication interface 902 can optionally include a standard wired interface, a wireless interface (such as WI-FI, a mobile communication interface, etc.), and can be controlled by the processor 901 to be used for receiving and transmitting data. The communication interface 902 can also be used for the transmission and interaction of internal data of the computer device. The memory 903 (Memory) is the memory device in the computer device, used to store programs and data. It can be understood that the memory 903 here can include both the built-in memory of the computer device and, of course, the extended memory supported by the computer device. The memory 903 provides a storage space, and the operating system of the computer device is stored in this storage space, which can include but is not limited to: Android system, iOS system, Windows Phone system, etc. The present application does not make any limitations in this regard.

[0243] The embodiments of the present application also provide a computer-readable storage medium (Memory). A computer-readable storage medium is a memory device in a computer device, used to store programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device, and of course also the extended storage medium supported by the computer device. The computer-readable storage medium provides a storage space, and this storage space stores the processing system of the computer device. And, in this storage space, there are also stored one or more instructions suitable for being loaded and executed by the processor 901, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory, or a non-volatile memory, such as at least one disk memory; optionally, it can also be at least one computer-readable storage medium located far from the aforementioned processor.

[0244] In one embodiment, the computer device can specifically be Figure 1 the network data analysis network element 103 shown in the figure. The processor 901 performs the following operations by running the executable program code in the memory 903:

[0245] Create a key pair, and send the first public key to N network elements in the network. The first public key is the public key in the key pair, and N is an integer greater than 1;

[0246] Obtain N first encrypted messages. The i-th first encrypted message is obtained by encrypting the gradient and loss information of the i-th network element using the first public key. The gradient and loss information of the i-th network element is calculated based on the first prediction data of this network element. The first prediction data is obtained by predicting the target sample parameters of the i-th network element through the model to be trained. The target sample parameters are the sample parameters obtained after aligning the sample parameters of the i-th network element with the sample parameters of N - 1 network elements. i is a positive integer less than or equal to N;

[0247] Based on the first private key and N first encrypted messages, generate parameter correction information. The parameter correction information includes the total gradient and total loss information of N network elements. The first private key is the private key in the key pair;

[0248] Send the parameter correction information to each network element, so that each network element updates the parameters of the model to be trained of this network element based on the parameter correction information.

[0249] As an optional embodiment, the processor 901 also performs the following operations by running the executable program code in the memory 903:

[0250] Obtain N second encrypted messages. The i-th second encrypted message is obtained by the i-th network element encrypting the second predicted data of the network element using the first public key. The second predicted data is obtained by predicting the target network parameters of the i-th network element using the i-th sub-model. The target network parameters are the network parameters obtained by aligning the network parameters of the i-th network element with the network parameters of N - 1 network elements.

[0251] Based on the first private key and N second encrypted messages, determine the service parameters associated with the target service.

[0252] Send the service parameters to the i-th network element so that the i-th network element adjusts the service strategy of the target service based on the service parameters.

[0253] As an optional embodiment, a specific embodiment in which the processor 901 determines the service parameters associated with the target service based on the first private key and N second encrypted messages is:

[0254] Decrypt the N second encrypted messages using the first private key to obtain N plaintext data.

[0255] Obtain M third encrypted messages. The third encrypted message is obtained by a network element in the network encrypting the network parameters of the network element using the second public key. The second public key is provided by the i-th network element, and M is an integer greater than 1.

[0256] Aggregate the M third encrypted messages to obtain an aggregation result corresponding to the M third encrypted messages.

[0257] Package the aggregation result and the N plaintext data to obtain the service parameters associated with the target service.

[0258] As an optional embodiment, a specific embodiment in which the processor 901 determines the service parameters associated with the target service based on the first private key and N second encrypted messages is:

[0259] Decrypt the N second encrypted messages using the first private key to obtain N plaintext data.

[0260] Perform integration processing on the N plaintext data to obtain the service parameters associated with the target service. The integration processing includes at least one of the following: data format conversion, data aggregation, data intersection, data filtering, data analysis.

[0261] As an optional embodiment, a specific embodiment in which the processor 901 generates parameter correction information based on the first private key and N first encrypted messages is:

[0262] Decrypt the N first encrypted messages using the first private key to obtain the gradients and loss information of the N network elements.

[0263] Sum the gradients of N network elements to obtain the total gradient of the N network elements;

[0264] Sum the loss information of N network elements to obtain the total loss information of the N network elements;

[0265] Package the total gradient and total loss information of N network elements to obtain parameter correction information.

[0266] As an alternative embodiment, the processor 901 further performs the following operations by running the executable program code in the memory 903:

[0267] Obtain a service subscription request, which carries the identities of N network elements participating in collaborative training and the identity information of the sender of the service subscription request;

[0268] Verify the identity information;

[0269] If the identity information passes the verification, return a response message to the sender of the service subscription request, and the response message is used to indicate that the service subscription request has passed.

[0270] As an alternative embodiment, the network elements in the network include user equipment, radio access network elements, and application function network elements;

[0271] The network parameters of the user equipment include at least one of the following: single-user signal-to-interference-plus-noise ratio, single-user received signal strength indication, single-user reference signal received power, single-user reference signal received quality, single-user delay, single-user data rate;

[0272] The network parameters of the radio access network elements include at least one of the following: signal-to-interference-plus-noise ratio, received signal strength indication, reference signal received power, reference signal received quality, delay, data rate, base station identifier;

[0273] The network parameters of the application function network elements are determined based on the target service.

[0274] In another embodiment, the computer device may specifically be Figure 1 The application function network element 102 shown. The processor 901 performs the following operations by running the executable program code in the memory 903:

[0275] Obtain a first public key and target sample parameters, where the target sample parameters are the sample parameters obtained by aligning the sample parameters of the application function network element with the sample parameters of N-1 network elements in the network, and N is an integer greater than 1;

[0276] Predict the target sample parameters through the model to be trained to obtain first prediction data;

[0277] Generate first encrypted information based on the first prediction data and the first public key, and return the first encrypted information to the provider of the first public key. The first encrypted information is obtained by encrypting the gradient and loss information of the application function network element, and the gradient and loss information of the application function network element are calculated based on the first prediction data;

[0278] Obtain parameter correction information sent by the provider of the first public key. The parameter correction information includes the total gradient and total loss information of N network elements, and the total gradient and total loss information of N network elements are calculated based on the gradient and loss information of the application function network element;

[0279] Update the parameters of the model to be trained based on the parameter correction information to obtain a risk prediction model.

[0280] As an alternative embodiment, the processor 901 also performs the following operations by running the executable program code in the memory 903:

[0281] Obtain target network parameters, where the target network parameters are network parameters obtained by aligning the network parameters of the application function network element with the network parameters of N-1 network elements in the network;

[0282] Use the risk prediction model to predict the target network parameters to obtain second prediction data;

[0283] Encrypt the second prediction data with the first public key to obtain second encrypted information, and return the second encrypted information to the provider of the first public key;

[0284] Obtain service parameters sent by the provider of the first public key, where the service parameters are associated with the target service;

[0285] Adjust the service strategy of the target service based on the service parameters.

[0286] As an alternative embodiment, the service parameters include an aggregation result and N plaintext data. The N plaintext data are obtained by decrypting the first encrypted information of N network elements, and the aggregation result is obtained by aggregating M third encrypted information. The jth third encrypted information is obtained by the jth network element encrypting the network parameters of the network element with the second public key provided by the application function network element, where j is a positive integer less than or equal to M;

[0287] A specific embodiment in which the processor 901 adjusts the service strategy of the target service based on the service parameters is:

[0288] Decrypt the aggregation result with the second private key corresponding to the second public key to obtain aggregation parameters;

[0289] Adjust the service strategy of the target service based on the aggregation parameters and the N plaintext data.

[0290] As an alternative embodiment, the target service is a remote control driving service, and the service parameters include the communication quality at the first moment. A specific embodiment in which the processor 901 adjusts the service strategy of the target service based on the service parameters is as follows:

[0291] If the communication quality at the first moment is lower than the quality threshold, then reduce the driving speed of the remotely controlled object at the second moment, where the second moment is before the first moment.

[0292] As an alternative embodiment, a specific embodiment in which the processor 901 generates the first encrypted information based on the first prediction data and the first public key is as follows:

[0293] Obtain N - 1 intermediate data sent by N - 1 network elements; the i-th intermediate data is obtained by the i-th network element predicting the target sample parameters of the network element using the model to be trained in the network element, or is obtained by the i-th network element aggregating the target sample parameters of the network element;

[0294] Based on the first prediction data and the N - 1 intermediate data, calculate the gradient and loss information;

[0295] Encrypt the gradient and loss information using the first public key to obtain the first encrypted information.

[0296] Based on the same inventive concept, the principle of problem - solving and the beneficial effects of the computer device provided in the embodiments of the present application are similar to those of the data - processing method in the method embodiments of the present application. One can refer to the principle of problem - solving and the beneficial effects of the method implementation. For the sake of brevity, they will not be elaborated here.

[0297] The embodiments of the present application further provide a computer - readable storage medium, in which a computer program is stored. The computer program is adapted to be loaded and executed by a processor to perform the data - processing method in the above - mentioned method embodiments.

[0298] The embodiments of the present application further provide a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer - readable storage medium. The processor of the computer device reads the computer instructions from the computer - readable storage medium, and the processor executes the computer instructions, so that the computer device performs the above - mentioned data - processing method.

[0299] The steps in the method embodiments of the present application can be adjusted, combined, and deleted according to actual needs.

[0300] The modules in the device embodiments of the present application can be combined, divided, and deleted according to actual needs.

[0301] In the embodiments of the present application, the "module" or "unit" refers to a computer program with a predetermined function or a part of a computer program, which works together with other related parts to achieve a predetermined goal, and can be fully or partially implemented by using software, hardware (such as a processing circuit or a memory), or a combination thereof. Similarly, one processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be a part of the overall module or unit that includes the function of the module or unit.

[0302] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. The readable storage medium can include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, an optical disc, etc.

[0303] The above disclosure is only a preferred embodiment of the present application. Of course, it cannot be used to limit the scope of rights of the present application. Those of ordinary skill in the art can understand all or part of the processes of implementing the above embodiments, and the equivalent changes made according to the claims of the present application still fall within the scope covered by the application.

Claims

1. A data processing method, characterized in that, The method includes: Create a key pair and send the first public key to N network elements in the network. The first public key is the public key in the key pair, and N is an integer greater than 1; Obtain N first encrypted messages. The i-th first encrypted message is obtained by encrypting the gradient and loss information of the i-th network element using the first public key. The gradient and loss information of the i-th network element is calculated based on the first prediction data of the network element. The first prediction data is obtained by predicting the target sample parameters of the i-th network element using a model to be trained. The target sample parameters are the sample parameters obtained by aligning the sample parameters of the i-th network element with the sample parameters of N - 1 network elements. i is a positive integer less than or equal to N; Generate parameter correction information based on the first private key and the N first encrypted messages. The parameter correction information includes the total gradient and total loss information of the N network elements. The first private key is the private key in the key pair; Send the parameter correction information to each network element so that each network element updates the parameters of the model to be trained of the network element based on the parameter correction information.

2. The method according to claim 1, characterized in that, The method further includes: Obtain N second encrypted messages. The i-th second encrypted message is obtained by the i-th network element encrypting the second prediction data of the network element using the first public key. The second prediction data is obtained by predicting the target network parameters of the i-th network element using the i-th sub-model. The target network parameters are the network parameters obtained by aligning the network parameters of the i-th network element with the network parameters of N - 1 network elements; Determine the service parameters associated with the target service based on the first private key and the N second encrypted messages; Send the service parameters to the i-th network element so that the i-th network element adjusts the service strategy of the target service based on the service parameters.

3. The method according to claim 2, characterized in that, The determining the service parameters associated with the target service based on the first private key and the N second encrypted messages includes: Decrypt the N second encrypted messages using the first private key to obtain N plaintext data; Obtain M third encrypted messages. The third encrypted message is obtained by a network element in the network encrypting the network parameters of the network element using a second public key. The second public key is provided by the i-th network element, and M is an integer greater than 1; Aggregate the M third encrypted messages to obtain an aggregation result corresponding to the M third encrypted messages; Package the aggregation result and the N plaintext data to obtain the service parameters associated with the target service.

4. The method according to claim 2, characterized in that, The determining the service parameters associated with the target service based on the first private key and the N second encrypted messages includes: Decrypt the N second encrypted messages using the first private key to obtain N plaintext data; Perform integration processing on the N plaintext data to obtain the service parameters associated with the target service. The integration processing includes at least one of the following: data format conversion, data aggregation, data intersection, data filtering, data analysis.

5. The method according to claim 1, characterized in that, The generating the parameter correction information based on the first private key and the N first encrypted messages includes: Decrypt the N first encrypted messages with the first private key to obtain the gradients and loss information of the N network elements; Sum up the gradients of the N network elements to obtain the total gradient of the N network elements; Sum up the loss information of the N network elements to obtain the total loss information of the N network elements; Package the total gradient and total loss information of the N network elements to obtain parameter correction information.

6. The method according to claim 1, characterized in that, The method further includes: Obtain a service subscription request, where the service subscription request carries the identifiers of N network elements participating in collaborative training and the identity information of the sender of the service subscription request; Verify the identity information; If the identity information passes the verification, return a response message to the sender of the service subscription request, where the response message is used to indicate that the service subscription request passes.

7. A data processing method, characterized in that, The method includes: Obtain a first public key and target sample parameters, where the target sample parameters are sample parameters obtained by aligning the sample parameters of the application function network element with the sample parameters of N - 1 network elements in the network, and N is an integer greater than 1; Predict the target sample parameters through the model to be trained to obtain first prediction data; Generate a first encrypted message based on the first prediction data and the first public key, and return the first encrypted message to the provider of the first public key, where the first encrypted message is an encrypted message of the gradients and loss information of the application function network element, and the gradients and loss information of the application function network element are calculated based on the first prediction data; Obtain the parameter correction information sent by the provider of the first public key, where the parameter correction information includes the total gradient and total loss information of N network elements, and the total gradient and total loss information of the N network elements are calculated based on the gradients and loss information of the application function network element; Update the parameters of the model to be trained based on the parameter correction information to obtain a risk prediction model.

8. The method according to claim 7, characterized in that, The method further includes: Obtain target network parameters, where the target network parameters are network parameters obtained by aligning the network parameters of the application function network element with the network parameters of N - 1 network elements in the network; Predict the target network parameters using the risk prediction model to obtain second prediction data; Encrypt the second prediction data with the first public key to obtain a second encrypted message, and return the second encrypted message to the provider of the first public key; Obtain service parameters sent by the provider of the first public key, where the service parameters are associated with a target service; Adjust the service strategy of the target service based on the service parameters.

9. The method according to claim 8, wherein, The service parameters include an aggregation result and N plaintext data, where the N plaintext data are obtained by decrypting the first encrypted messages of the N network elements, and the aggregation result is an aggregation of M third encrypted messages. The j-th third encrypted message is obtained by the j-th network element encrypting its network parameters with the second public key provided by the application function network element, and j is a positive integer less than or equal to M; The adjusting the service strategy of the target service based on the service parameters includes: Decrypt the aggregation result using the second private key corresponding to the second public key to obtain aggregation parameters; Adjust the service strategy of the target service based on the aggregation parameters and the N plaintext data.

10. The method according to claim 8, wherein, The target service is a remote control driving service, and the service parameters include the communication quality at the first moment; adjusting the service strategy of the target service based on the service parameters includes: If the communication quality at the first moment is lower than the quality threshold, reduce the driving speed of the object to be remotely controlled at the second moment, where the second moment is before the first moment.

11. The method according to claim 7, wherein, Generating the first encrypted information based on the first prediction data and the first public key includes: Obtain N - 1 intermediate data sent by N - 1 network elements; the i-th intermediate data is obtained by the i-th network element using the model to be trained in the network element to predict the target sample parameters of the network element, or is obtained by aggregating the target sample parameters of the i-th network element by the i-th network element; Calculate the gradient and loss information based on the first prediction data and the N - 1 intermediate data; Encrypt the gradient and loss information using the first public key to obtain the first encrypted information.

12. A data processing device, wherein, The data processing device includes: A processing unit for creating a key pair and sending the first public key, which is the public key in the key pair, to N network elements in the network through a sending unit, where N is an integer greater than 1; An obtaining unit for obtaining N first encrypted information, where the i-th first encrypted information is obtained by encrypting the gradient and loss information of the i-th network element using the first public key, and the gradient and loss information of the i-th network element is calculated based on the first prediction data of the network element, and the first prediction data is obtained by predicting the target sample parameters of the i-th network element through the model to be trained, and the target sample parameters are the sample parameters obtained by aligning the sample parameters of the i-th network element with the sample parameters of N - 1 network elements, and i is a positive integer less than or equal to N; The processing unit is further configured to generate parameter correction information based on the first private key and the N first encrypted information, where the parameter correction information includes the total gradient and total loss information of the N network elements, and the first private key is the private key in the key pair; The sending unit is further configured to send the parameter correction information to each network element so that each network element updates the parameters of the model to be trained in the network element based on the parameter correction information.

13. A data processing device, wherein, The data processing device includes: An obtaining unit for obtaining the first public key and the target sample parameters, where the target sample parameters are the sample parameters obtained by aligning the sample parameters of the application function network element with the sample parameters of N - 1 network elements in the network, and N is an integer greater than 1; A processing unit for predicting the target sample parameters through the model to be trained to obtain the first prediction data; and for generating first encrypted information based on the first prediction data and the first public key, and returning the first encrypted information to the provider of the first public key through a sending unit, where the first encrypted information is obtained by encrypting the gradient and loss information of the application function network element, and the gradient and loss information of the application function network element is calculated based on the first prediction data; The obtaining unit is further configured to obtain parameter correction information sent by the provider of the first public key, where the parameter correction information includes the total gradient and total loss information of N network elements, and the total gradient and total loss information of the N network elements is calculated based on the gradient and loss information of the application function network element; The processing unit is further configured to update the parameters of the to-be-trained model based on the parameter correction information to obtain a risk prediction model.

14. A computer device, wherein, including: a memory, in which a computer program is stored; a processor, configured to load the computer program to implement the data processing method according to any one of claims 1-6, or configured to load the computer program to implement the data processing method according to any one of claims 7-11.

15. A computer-readable storage medium, wherein, The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded and executed by the processor to implement the data processing method according to any one of claims 1-6, or suitable for being loaded and executed by the processor to implement the data processing method according to any one of claims 7-11.

Images