Information transmission method and communication apparatus

Abstract

Provided in the embodiments of the present application are an information transmission method and a communication apparatus. The method comprises: a first network element sending a first request to a second network element, and receiving a first response from the second network element. The first request comprises first information, the first information being used for indicating at least one first sample alignment algorithm supported by the first network element. The first response is used for indicating a second sample alignment algorithm determined by the second network element from the at least one first sample alignment algorithm, the second sample alignment algorithm being used for determining a common sample between a data set used of the first network element used for training a vertical federated learning model and a data set of the second network element used for training a vertical federated learning model; or the first response is used for indicating not supporting executing the sample alignment. Therefore, by means of the first request and the first response, the first network element and the second network element can negotiate the sample alignment algorithm used in a sample alignment process, thereby improving the efficiency and success rate in the sample alignment between the first network element and the second network element.

Description

Information transmission method and communication device

[0001] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office on January 5, 2024, with application number 202410022999.3 and application name “Information Transmission Method and Communication Device”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of communications, and in particular to an information transmission method and a communication device. Background Art

[0003] Functional network elements in communication networks can leverage artificial intelligence (AI) technology and big data from communication networks to output analytical data through model training and inference, assisting in communication network policy formulation and network resource adjustment. However, when functional network elements acquire data sets, data privacy protections between different domains of the communication network may prevent them from obtaining the required data, resulting in infeasible model training or low accuracy.

[0004] To this end, distributed federated modeling can be used for model training to break down data silos. For example, vertical federated learning (VFL) can be used between multiple functional network elements (or participants) to jointly model without sharing raw data, enabling AI collaboration.

[0005] Currently, VFL requires that all participating datasets share the same sample space (or sample identifier set). This requires that all participating datasets be aligned before model training begins to determine the shared sample identifiers. However, specific approaches to sample alignment, to improve its efficiency and success rate, remain to be explored. Summary of the Invention

[0006] The information transmission method and communication device provided in the embodiments of the present application can improve the efficiency and success rate of sample alignment.

[0007] To achieve the above objectives, the embodiments of the present application adopt the following technical solutions:

[0008] In a first aspect, a method for information transmission is provided. The method can be executed by a component of a first network element, such as a processor, chip, or chip system of the first network element, and can also be implemented by a logic module or software that can implement all or part of the first network element. The method includes: the first network element sends a first request to a second network element, and receives a first response from the second network element. The first request includes first information, and the first information is used to indicate at least one first sample alignment algorithm supported by the first network element. The first response is used to indicate a second sample alignment algorithm among the at least one first sample alignment algorithm, and the second sample alignment algorithm is used to determine common samples between a dataset used for longitudinal federated learning model training of the first network element and a dataset used for longitudinal federated learning model training of the second network element; alternatively, the first response is used to indicate that sample alignment is not supported.

[0009] In the embodiment of the present application, the first network element sends a first request to the second network element, which enables the second network element to determine whether the second network element and the first network element support the same sample alignment algorithm to perform sample alignment based on at least one first sample alignment algorithm. If it is determined that the second network element and the first network element support the same sample alignment algorithm, the second network element can feedback the second sample alignment algorithm to the first network element through a first response. The first network element and the second network element can then negotiate the sample alignment algorithm used in the sample alignment process, thereby improving the efficiency and success rate of sample alignment between the first network element and the second network element. Furthermore, if the second network element does not support at least one first sample alignment algorithm, the second network element can feedback to the first network element through the first response that it does not support sample alignment, so that the first network element can promptly perform sample alignment with other participants participating in the vertical federated learning model training, thereby improving the efficiency of the first network element in performing sample alignment. Therefore, based on the information transmission method provided in the embodiment of the present application, the efficiency and success rate of sample alignment performed by the first network element can be improved.

[0010] In one possible implementation, the first network element may be a participant in vertical federated learning. For example, the first network element is the master participant in vertical federated learning model training, and the second network element is a candidate slave participant in vertical federated learning model training. In another example, the first network element is a slave participant in vertical federated learning model training, and the second network element is a candidate slave participant or master participant in vertical federated learning model training. In other words, either the master or the slave participant can initiate the sample alignment negotiation process, i.e., send a first request. It should be understood that, without distinguishing between master and slave participants, both the first network element and the second network element are participants in vertical federated learning model training. Furthermore, since participants in vertical federated learning model training can be network function network elements or terminal devices, the first network element can be a network function network element, a terminal device, or an access network device. For example, the first network element can be a network data analysis function network element, an application function network element, an access and mobility management function network element, a session management function network element, or a policy control function network element. Furthermore, similar to the first network element, the second network element can be a network function network element, a terminal device, or an access network device.

[0011] In a possible implementation, the method provided in the first aspect further includes: the first network element receives information of the second network element from the third network element, and the information of the second network element includes an identifier of the second network element and / or address information of the second network element. It can be understood that the third network element can be a trusted third party, for example, the third network element can be a trusted coordinator (such as a VFL server); or, the third network element can be a vertical federated learning support function (VFL support function, VFLSF) network element, and the third network element is used to store the sample alignment capabilities of the participants in the vertical federated learning model training. That is, the first network element can obtain the identifier of the second network element and / or the address information of the second network element by receiving information of the second network element from the third network element, so that the first network element can send a first request to the second network element.

[0012] In one possible implementation, the method provided in the first aspect further includes: the first network element sends a second request to the third network element, the second request includes second information, and the second information is used to indicate the sample alignment requirement. Accordingly, the first network element receives information from the second network element, including: the first network element receives a second response from the third network element, and the second response includes information of the second network element. In other words, the second network element can be a network element whose sample alignment capability meets the sample alignment requirement corresponding to the first network element, thereby reducing the probability that the second network element does not support sample alignment, thereby improving the success rate and efficiency of sample alignment negotiation between the first network element and the second network element, and further improving the success rate and efficiency of sample alignment.

[0013] In one possible implementation, the sample alignment requirement includes at least one of the following: a third sample alignment type, a third sample alignment algorithm, or a third time period for sample alignment. That is, based on one or more of the third sample alignment type, the third sample alignment algorithm, and the third time period for sample alignment included in the sample alignment requirement, a second network element that better meets the sample alignment requirement corresponding to the first network element can be screened, thereby further reducing the probability that the second network element does not support sample alignment, thereby further improving the success rate and efficiency of sample alignment negotiation between the first network element and the second network element.

[0014] In one possible implementation, the information about the second network element further includes at least one of the following: a fourth sample alignment type, a fourth sample alignment algorithm, or a fourth time period supporting sample alignment. That is, the first network element may further obtain the sample alignment capability of the second network element, so that the first network element can determine the sample alignment algorithm and time period commonly supported by the first network element and the second network element. This allows the primary participant to send the first request to the second network element based on the commonly supported sample alignment algorithm and / or commonly supported time period, thereby enabling more efficient sample alignment negotiation with the second network element.

[0015] In one possible implementation, the method provided in the first aspect further includes: the first network element sends a registration request to the third network element, the registration request includes third information, and the third information is used to indicate at least one of the following: the sample alignment type supported by the first network element, the sample alignment algorithm supported, or the time period for supporting sample alignment. In other words, the first network element can register the sample alignment capability with the third network element so that the third network element can obtain the sample alignment capability of the first network element (for example, the supported sample alignment type, the supported sample alignment algorithm, or the time period for supporting sample alignment, etc.), and select the second network element that meets the sample alignment requirements corresponding to the first network element based on the sample alignment capability of the first network element.

[0016] In one possible implementation, the at least one first sample alignment algorithm is an algorithm supported by both the first network element and the second network element. That is, the first network element can indicate to the second network element, through the first request, the at least one first sample alignment algorithm supported by both the first network element and the second network element. This can reduce the probability that the second network element does not support sample alignment, thereby improving the success rate and efficiency of sample alignment negotiation.

[0017] In one possible implementation, the first information is further used to indicate at least one of the following: the number of samples supported by the first network element for sample alignment, the first sample alignment type supported by the first network element, or at least one first time period for which the first network element supports sample alignment. That is, by indicating to the second network element through the first information the number of samples supported by the first network element for sample alignment, the first sample alignment type, or the at least one first time period, this can be used to assist the second network element in further determining whether to support sample alignment, thereby avoiding sample alignment failure during the actual execution of sample alignment by the second network element due to local resources not meeting the computing power requirements of the sample alignment algorithm, thereby improving the success rate and efficiency of sample alignment. For example, the number of samples supported by the first network element for sample alignment can refer to: the number of all samples in the dataset used by the first network element for training a longitudinal federated learning model, or a subset of the number of all samples in the dataset. The subset of the number of all samples can be used to indicate that the first network element expects to use the number of samples corresponding to the subset for sample alignment. In addition, the number of samples supported by the first network element for sample alignment can be used by the second network element to determine the computing power requirements of the sample alignment algorithm, and then the second network element can determine whether local resources can support the use of the first sample alignment algorithm to perform sample alignment during the sample alignment process, thereby improving the success rate and efficiency of the sample alignment. For another example, the sample alignment type supported by the first network element can be a sample alignment type that the first network element can support, and can also be used to indicate the sample alignment type that the first network element expects to perform sample alignment. The sample alignment type may support two-party sample alignment and / or multi-party sample alignment. For a sample alignment algorithm that can simultaneously support two-party sample alignment and multi-party sample alignment, the computing power requirements of the sample alignment algorithm for two-party sample alignment and the computing power requirements for multi-party sample alignment may be different. Then, the sample alignment type supported by the first network element can assist the second network element in determining the computing power requirements of the sample alignment algorithm, so that the second network element can determine whether to support sample alignment, thereby improving the success rate and efficiency of sample alignment. For another example, the first network element supports at least one time period for sample alignment, which can be used by the second network element to determine whether to support sample alignment. The second network element can determine whether to support sample alignment based on the at least one time period and the load conditions within the at least one time period, thereby improving the success rate and efficiency of sample alignment.

[0018] In one possible implementation, at least one first sample alignment algorithm is associated with at least one first time period. That is, by associating at least one first sample alignment algorithm with at least one first time period, a second network element can determine which first sample alignment algorithms the first network element supports in different first time periods, and determine whether computing power requirements corresponding to the first sample alignment algorithms in different first time periods are supported based on available local resources in the different first time periods. This allows the second network element to further determine whether sample alignment is supported, thereby improving the success rate and efficiency of sample alignment.

[0019] In one possible implementation, the first response is further used to indicate the number of samples supported by the second network element for sample alignment and / or the second time period for which the second network element supports sample alignment; or the first response is further used to indicate the reason why the second network element does not support sample alignment and / or the maximum number of samples that the second network element expects the first network element to use. That is, if the second network element supports sample alignment, the first response further indicating the number of samples supported by the second network element for sample alignment and / or the second time period for which the second network element supports sample alignment can assist the first network element in further determining whether sample alignment can be supported during the sample alignment process, thereby improving the success rate of sample alignment. In addition, if the second network element does not support sample alignment, the first response further indicating the reason why the second network element does not support sample alignment and / or the maximum number of samples that the second network element expects the first network element to use can assist the first network element in adjusting parameters such as the first sample alignment algorithm indicated in the first information, the number of samples supported by the first network element for sample alignment, or the at least one first time period for which the first network element supports sample alignment when subsequently sending the first request, thereby improving the success rate of sample alignment negotiation.

[0020] In one possible implementation, the method provided in the first aspect further includes: the first network element sends a first message to the second network element, where the first message is used to indicate a target sample alignment algorithm, where the target sample alignment algorithm is a sample alignment algorithm determined from at least one first sample alignment algorithm based on the first response. In other words, the first network element can further determine, based on the sample alignment response message from the second network element, that the first network element and the second network element can support sample alignment, as well as the target sample alignment algorithm used for sample alignment, to avoid sample alignment failure due to failure to meet the computing power requirements of the sample alignment algorithm during the sample alignment process, thereby improving the success rate and efficiency of sample alignment.

[0021] In one possible implementation, the method provided in the first aspect further includes: the first network element sends a first request to the fourth network element; the first network element receives a third response from the fourth network element, the third response is used to indicate a fifth sample alignment algorithm in at least one first sample alignment algorithm, and the fifth sample alignment algorithm is used to determine the common samples between the data set used for longitudinal federated learning model training of the first network element and the data set used for longitudinal federated learning model training of the fourth network element; the first network element sends a second message to the fourth network element, and the second message is used to indicate a target sample alignment algorithm; wherein, there are multiple fifth sample alignment algorithms and / or second sample alignment algorithms, and the target sample alignment algorithm is a sample alignment algorithm determined from at least one first sample alignment algorithm based on the second sample alignment algorithm, including: the target sample alignment algorithm is a sample alignment algorithm determined from at least one first sample alignment algorithm based on the second sample alignment algorithm and the fifth sample alignment algorithm.

[0022] That is to say, the first network element can send a first request to multiple network elements to obtain the sample alignment algorithm selected by multiple network elements from at least one sample alignment algorithm, and then determine the target sample alignment algorithm that can be jointly performed by the first network element and multiple network elements, thereby realizing multi-party sample alignment.

[0023] In one possible implementation, the first message is further used to indicate a target time period for sample alignment. That is, the first network element may also indicate the target time period for performing sample alignment via the first message, thereby notifying the second network element to perform the sample alignment process within the target time period. This prevents the second network element from performing the sample alignment process during a time period when the second network element may not be able to meet the computing power requirements of the target sample alignment algorithm, thereby further improving the success rate and efficiency of sample alignment.

[0024] In one possible implementation, the method provided in the first aspect further includes: when the first network element determines not to perform sample alignment, the first network element sending a third message to the second network element, where the third message is used to indicate that sample alignment is not to be performed. In other words, by sending the third message to the second network element to indicate that sample alignment is not to be performed, the first network element can promptly notify the second network element that sample alignment will not be performed next time.

[0025] In one possible implementation, the third message is further used to indicate the reason for not performing sample alignment. That is, by indicating the reason for not performing sample alignment to the second network element, the first network element can assist the second network element in adjusting the second sample alignment algorithm, the number of samples supported for sample alignment, or the time period supported for sample alignment indicated in the first response, when the first network element subsequently sends the first request again, thereby improving the success rate and efficiency of sample alignment negotiation.

[0026] In a second aspect, a method for information transmission is provided. The method can be executed by a component of a first network element, such as a processor, chip, or chip system of the first network element, or can be implemented by a logic module or software that can implement all or part of the first network element. The method includes: the first network element sends a first request to a second network element, and receives a first response from the second network element. The first request includes first indication information, and the first indication information is used to indicate the provision of a sample identifier for training a vertical federated learning model. The first response includes the first sample identifier.

[0027] In the embodiment of the present application, the first network element can send a first request to the second network element, causing the second network element to feedback a first sample identifier to the first network element. The first network element can then determine, based on the first sample identifier, the common sample identifiers among the various participants participating in the vertical federated learning model training, thereby achieving sample alignment. For example, the first network element may be a trusted third party, and the second network element may be a master or slave participant in the vertical federated learning. The first network element can then send a first request to the second network element to obtain the sample identifiers corresponding to the second network element's dataset used for vertical federated learning model training, so that the first network element can determine the common samples among the various participants participating in the same vertical federated learning model training, thereby achieving sample alignment. For another example, the second network element may be a trusted third party for registering sample identifiers corresponding to the participant's dataset, and the first network element may be a master, slave, or coordinator of the vertical federated learning model training. The second network element can then, in response to the first request from the first network element, feedback the common sample identifiers of at least two network elements participating in the same vertical federated learning model training to the first network element, thereby achieving sample alignment.

[0028] In one possible implementation, the first indication information is used to indicate the provision of a sample identifier for longitudinal federated learning model training, including: the first indication information is used to indicate the provision of a sample identifier corresponding to a data set for longitudinal federated learning model training, and the first sample identifier is a sample identifier corresponding to a data set for longitudinal federated learning model training from a second network element.

[0029] That is to say, the first network element can provide the sample identifier corresponding to its data set used for longitudinal federated learning model training to the second network element through the sample provision indication in the sample alignment request message, so that the second network element can feedback the sample identifier corresponding to the second network element's data set used for longitudinal federated learning model training to the first network element, and then trigger the first network element to determine the common sample identifier between the second network element and other candidate participants participating in the longitudinal federated learning model training, thereby achieving sample alignment.

[0030] In one possible implementation, the second network element is a candidate network element for participating in longitudinal federated learning model training. The method provided in the second aspect further includes: the first network element receiving a second sample identifier corresponding to a dataset for longitudinal federated learning model training from a third network element, and sending a third message to the second network element based on the first sample identifier and the second sample identifier. The third network element is a network element other than the second network element among multiple candidate network elements participating in longitudinal federated learning model training. The third message includes second indication information, the second indication information being used to instruct the second network element to participate in longitudinal federated learning model training, and / or a third sample identifier, the third sample identifier being a common sample identifier of at least two network elements participating in longitudinal federated learning model training, the at least two network elements being at least two network elements determined from multiple candidate network elements based on the first sample identifier and the second sample identifier, the at least two network elements including the third network element; or the second indication information being used to indicate that the second network element does not participate in longitudinal federated learning model training, and / or the reason why the second network element does not participate in longitudinal federated learning model training. That is to say, after the first network element obtains the sample identifier corresponding to the data set for vertical federated learning model training provided by at least some of the candidate network elements among the multiple candidate network elements, it can further determine at least two network elements that can participate in the vertical federated learning model training from the multiple candidate network elements based on the first sample identifier of the second network element and the second sample identifier of the third network element, as well as the common sample identifier (i.e., the third sample identifier) ​​of the at least two network elements, thereby achieving sample alignment and increasing the probability of the second network element participating in the vertical federated learning model training. In addition, if the first network element determines that the second network element does not participate in the vertical federated learning model training, the first network element sends a third message to the second network element to instruct the second network element not to participate in the vertical federated learning model training, so that the second network element can promptly determine not to participate in this vertical federated learning model training, so that the second network element can prepare to execute other related processes of vertical federated learning. Furthermore, by indicating the reason why the second network element does not participate in the vertical federated learning model training through the third message, it can be convenient for the second network element to participate in the vertical federated learning model training again, and it can assist the second network element to adjust the sample identifier corresponding to the data set provided for the vertical federated learning model training, so as to increase the probability of the second network element participating in the vertical federated learning model training.

[0031] In one possible implementation, the second network element is a network element for registering sample identifiers corresponding to data sets of network elements participating in longitudinal federated learning model training; the first indication information is used to indicate the provision of sample identifiers for longitudinal federated learning model training, including: the first indication information is used to indicate the provision of a common sample identifier, the common sample identifier is a common sample identifier between data sets of at least two network elements participating in the same longitudinal federated learning model training among multiple network elements, and the first sample identifier is used to indicate a common sample between data sets of at least two network elements participating in the first longitudinal federated learning model training among multiple network elements. That is, for the second network element being a trusted third party for registering sample identifiers corresponding to network element data sets participating in longitudinal federated learning model training, the first network element can send a first request to the second network element so that the second network element can feedback to the first network element the common sample identifier between data sets of at least two network elements participating in the same longitudinal federated learning model training among multiple network elements, thereby achieving sample alignment.

[0032] In one possible implementation, the first network element is a network element participating in the training of a longitudinal federated learning model; the method provided in the second aspect further includes: the first network element sending a registration request to the second network element, the registration request including fourth indication information, the fourth indication information being used to indicate a sample identifier corresponding to a data set of the first network element used for the training of the longitudinal federated learning model. In other words, the first network element may send a registration request to the second network element to register with the second network element the sample identifier corresponding to the data set of the first network element used for the training of the longitudinal federated learning model, so that the second network element can determine the common sample identifier between the first network element and at least one participant participating in the same longitudinal federated learning model training as the first network element, thereby achieving sample alignment.

[0033] In one possible implementation, the first request also includes fifth indication information and / or sixth indication information, the fifth indication information is used to indicate the number range of sample identifiers contained in the common sample identifier, and the sixth indication information is further used to instruct the second network element to determine at least two network elements. That is, by indicating the number range of samples indicated by the common sample identifier through the fifth indication information, it is possible to avoid the second network element from determining too many or too few common samples, thereby increasing the probability that the participants corresponding to the common sample identifier will join the vertical federated learning model training. In addition, in the case where the first request is not used to request the discovery of at least two network elements participating in the same vertical federated learning model training among multiple network elements, the sixth indication information can be used to instruct the second network element to determine at least two network elements participating in the same vertical federated learning model training among multiple network elements.

[0034] In one possible implementation, the first response also includes the identifiers and / or address information of at least two network elements, and / or the number of sample identifiers contained in the first sample identifier. That is, the second network element feeds back the identifiers and / or address information of at least two network elements to the first network element through the first response, so that the first network element can notify the common sample identifier corresponding to the at least two network elements, thereby completing the sample alignment. In addition, in the case where the first network element is a VFL server and the encryption algorithm or decryption algorithm corresponding to the first sample identifier is not configured, the second network element indicates the number of sample identifiers contained in the first sample identifier to the first network element through the first response, so that the first network element can determine the first sample identifier within the common sample quantity range that meets the training requirements of the vertical federated learning model, thereby increasing the probability of the participants joining the vertical federated learning model training.

[0035] In a third aspect, a communication device is provided for implementing the various methods described above. The communication device may be the first network element in any of the above aspects or any of its implementations, or a device comprising the first network element, or a device contained in the first network element, such as a chip. The communication device includes modules, units, or means corresponding to the implementation of the above methods. The modules, units, or means may be implemented by hardware, software, or by hardware executing corresponding software implementations. The hardware or software includes one or more modules or units corresponding to the above functions.

[0036] In some possible designs, the communication device may include a processing module and a transceiver module. The transceiver module, also referred to as a transceiver unit, is configured to implement the transmitting and / or receiving functions described in any of the above aspects and any possible implementations thereof. The transceiver module may be comprised of a transceiver circuit, a transceiver, a transceiver, or a communication interface. The processing module may be configured to implement the processing functions described in any of the above aspects and any possible implementations thereof.

[0037] In some possible designs, the transceiver module includes a sending module and a receiving module, which are respectively used to implement the sending and receiving functions in any of the above aspects and any possible implementation methods.

[0038] In a fourth aspect, a communication device is provided, comprising: at least one processor; the processor is configured to execute a computer program or instruction so that the communication device executes the method described in any one of the above aspects.

[0039] In one possible implementation, the communication device further includes the memory. Optionally, the memory is coupled to the processor, the memory may be integrated with the processor, or the memory may be independent of the processor. Optionally, the processor is configured to execute computer programs or instructions stored in the memory.

[0040] In a possible implementation, the memory is independent of the communication device.

[0041] In a possible implementation, the communication device further includes a communication interface, which is used to communicate with a module outside the communication device.

[0042] The communication device may be the first network element in any of the above aspects or any of its implementations, or a device including the above first network element, or a device included in the above first network element, such as a chip.

[0043] In a fifth aspect, a computer-readable storage medium is provided, which stores a computer program or instruction. When the computer-readable storage medium is run on a communication device, the communication device can execute the method described in any of the above aspects or any of its implementation methods.

[0044] In a sixth aspect, a computer program product comprising instructions is provided, which, when executed on a communication device, enables the communication device to execute the method described in any one of the above aspects or any one of its implementations.

[0045] In a seventh aspect, a communication device is provided (for example, the communication device may be a chip or a chip system), which includes a processor for implementing the functions involved in any of the above aspects or any of its implementation methods.

[0046] In some possible designs, the communication device includes a memory for storing necessary program instructions and data.

[0047] In some possible designs, when the device is a chip system, it can be composed of a chip or include a chip and other discrete devices.

[0048] It can be understood that when the communication device provided in any one of the third to seventh aspects is a chip, the above-mentioned sending action / function can be understood as output, and the above-mentioned receiving action / function can be understood as input.

[0049] Among them, the technical effects brought about by any design method in the third to seventh aspects can refer to the technical effects brought about by the different design methods in the above-mentioned first aspect, and will not be repeated here.

[0050] In an eighth aspect, a communication system is provided, comprising: a first network element and a second network element in any of the above aspects or any of its implementations, the second network element being configured to receive a first request and send a first response. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] FIG1 is a schematic diagram of feature distribution of a dataset for vertical federated learning provided in an embodiment of the present application;

[0052] FIG2 is a schematic diagram of participants in vertical federated learning provided by an embodiment of the present application;

[0053] FIG3 is a schematic diagram of a vertical federated learning model training process provided in an embodiment of the present application;

[0054] FIG4 is a schematic diagram of two-way sample alignment provided by an embodiment of the present application;

[0055] FIG5 is a schematic diagram of the architecture of a communication system provided in an embodiment of the present application;

[0056] FIG6 is a flowchart of an information transmission method according to an embodiment of the present application;

[0057] FIG7 is a second flow chart of an information transmission method provided in an embodiment of the present application;

[0058] FIG8 is a third flow chart of an information transmission method provided in an embodiment of the present application;

[0059] FIG9 is a fourth flow chart of an information transmission method provided in an embodiment of the present application;

[0060] FIG10 is a flowchart of an information transmission method according to an embodiment of the present application;

[0061] FIG11 is a sixth flow chart of an information transmission method according to an embodiment of the present application;

[0062] FIG12 is a flow chart of an information transmission method according to an embodiment of the present application;

[0063] FIG13 is a structural diagram of a communication device according to an embodiment of the present application;

[0064] FIG14 is a second structural diagram of a communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0065] To facilitate understanding of the technical solutions provided by the embodiments of this application, a brief introduction to the relevant technical terms of this application is first given. The brief introduction is as follows:

[0066] First, an intelligent network architecture based on the network data analytics function (NWDAF):

[0067] Currently, the Third Generation Partnership Project (3GPP) has defined an intelligent network architecture based on NWDAF. Its purpose is to collect massive amounts of information from the network and utilize existing big data and AI technologies to utilize this data. The output of valuable information can assist operators in policy formulation and network resource adjustment, thereby improving user experience and reducing network load.

[0068] For example, Table 1 lists some analysis results that the NWDAF network element can provide, as well as the data that the NWDAF network element needs to collect to provide the corresponding analysis results. For example, the NWDAF network element can provide service experience analysis results. In order to provide the analysis results, the NWDAF network element can collect service identification (ID) or service experience and other service-related information from the application function (AF) network element or terminal device (or user equipment (UE)), and collect signal reception power and signal reception quality and other information from operations, administration and management (OAM). The NWDAF can train an AI model based on the collected data, and then obtain inference analysis results based on the AI ​​model, such as obtaining a predicted service experience for a certain time period in the future.

[0069] Table 1

[0070] Terminal devices, core networks (CN), or applications (APP) within the network may not report terminal device or application-related data to NWDAF due to data privacy protection issues, resulting in NWDAF being unable to obtain the required data, making model training unfeasible or significantly reducing accuracy.

[0071] To solve the above problems, federated learning (FL) can be used for model training.

[0072] Second, federated learning:

[0073] A machine learning (ML) framework can effectively help multiple organizations utilize data and conduct machine learning modeling while meeting user privacy, data security, and government regulations. Federated learning, as a distributed ML framework, can effectively address data silos, allowing participants to jointly model without sharing raw data. This technically breaks down data silos and enables AI collaboration.

[0074] Federated learning can be divided into three categories based on the feature distribution between the datasets used by each participant to train the federated learning model: horizontal federated learning, vertical federated learning, and federated transfer learning. The feature distribution between the datasets of each participant primarily refers to the distribution of the sample space and feature space between the datasets of each participant.

[0075] It should be understood that a dataset can correspond to two dimensions: one dimension is the sample space, which can be used to represent (or characterize): the set of objects reflected or described by the dataset; the other dimension is the feature space, which can be used to represent (or characterize): the set of features corresponding to the data reflected or described by the dataset. For example, an object can be a user, account, terminal device, or application. A feature can be an attribute of an object, such as location, signal reception quality, or speed.

[0076] Exemplarily, the object reflected or described by the data set is a group of UEs, that is, the sample space is the group of UEs, each UE in the group of UEs can be called a sample, and different UEs in the group of UEs can be distinguished by a sample identifier. The characteristics corresponding to the data of the object reflected or described by the data set can be the position, signal reception power, or speed of the UE in Table 1. For example, referring to Table 2, the data set of the participant can be composed of data of three samples, and the three samples are UE1, UE2, and UE3 respectively. The data of the three samples can reflect or describe the location characteristics and signal reception power characteristics of the UE. Among them, the sample identifier of sample UE1 is U1, and the data of sample UE1 includes location data #1 and signal reception power data #1. The sample identifier of sample UE2 is U2, and the data of sample UE2 includes location data #2 and signal reception power data #2. The sample identifier of sample UE3 is U3, and the data of sample UE3 includes location data #3 and signal reception power data #3.

[0077] In addition, the sample identifier can be an identifier of the UE, such as a user hidden identifier (SUCI) of the UE, a user permanent identifier (SUPI), or an international mobile subscriber identity (IMSI), etc., or it can be other identifiers that can be used to represent the identity of the UE. The embodiment of the present application does not specifically limit this.

[0078] It can be understood that since the sample identifier is the SUCI or SUPI of the UE, the sample identifier involves the user's data privacy.

[0079] Table 2

[0080] It should be understood that in the embodiments of the present application, samples can also be understood as the row dimension in the above Table 2. In addition, Table 2 is only an example, and samples can also be accounts, applications, sensing targets (such as vehicles or transportation infrastructure), or images, etc., which are not specifically limited in the embodiments of the present application.

[0081] Based on the above descriptions of datasets, sample spaces, and feature spaces, the following example illustrates the feature distribution between datasets of various participants in vertical federated learning.

[0082] Third, vertical federated learning:

[0083] For vertical federated learning, its main characteristic distribution is that the sample space overlap between the data sets of each participant is high, and the feature space overlap is low.

[0084] Figure 1 is a feature distribution diagram of a dataset for vertical federated learning provided in an embodiment of the present application. As shown in Figure 1, the dataset of participant A contains the data of [user 1, user 2, user 3, user 4], and the data of each user contains [feature 1, feature 2, feature 3, feature 4, feature 5]. The dataset of participant B contains the data of [user 1, user 2, user 3, user 4, user 5], and the data of each user contains [feature 5, feature 6, feature 7, feature 8, feature 9]. It can be seen that the data feature intersection between the datasets of participant A and participant B is very small, but most of the samples are the same. Participant A and participant B can select the same samples (i.e., common samples (users 1 to 4)) from their respective datasets to perform model training for vertical federated learning, so that the accuracy of the model can be improved by expanding the number of features.

[0085] It can be understood that Figure 1 above is only an example. The samples between the data set of participant A and the data set of participant B can be exactly the same (for example, both include users 1 to 4), and the data features between the data set of participant A and the data set of participant B may have no intersection. For example, the features contained in each sample in the data set of participant A are [features 1 to 5], and the data features contained in each sample in the data set of participant B are [features 6 to 10].

[0086] The following introduces the participants, model training, and sample alignment of vertical federated learning.

[0087] 3.1. For participants in vertical federated learning:

[0088] FIG2 is a schematic diagram of participants in a vertical federated learning provided in an embodiment of the present application. As shown in FIG2 , participants in vertical federated learning can be divided into three types: a master participant, a slave participant, and a trusted collaborator. Among them, a master participant may refer to: for a specific federated learning task, a participant who has a data label corresponding to the task, and the data label corresponding to the task is the result to be predicted or output by the task. For example, the federated learning task may be a business experience prediction, and the data label may be a business experience, that is, the data output by the federated learning task is business experience data.

[0089] A slave participant refers to a participant that possesses the required data features for a specific federated learning task, i.e., the input data required by the model for that task. For example, if the data label is UE service experience, the required data features for this task may include location, average throughput, or average packet delay.

[0090] A trusted collaborator (or trusted coordinator) can be responsible for maintaining the vertical federated learning (VFL) process, authorizing the entry and removal of federated learning members, and distributing encryption keys and decrypting intermediate information. The coordinator is typically a third party independent of the federated learning task or a credible institution. For example, the trusted coordinator can be the vertical federated learning server (VFL server).

[0091] It should be understood that the master participant or slave participant participating in the vertical federated learning task mentioned above can also be called a vertical federated learning client (VFL client).

[0092] In one possible implementation, the trusted coordinator can distribute the vertical federated learning model to multiple participants in the vertical federated learning task. The multiple participants perform model training based on the local vertical federated learning model and upload the intermediate parameters of the model training to the trusted coordinator. The trusted coordinator determines whether to terminate the model training process and distributes the decrypted intermediate parameters, etc., to realize the vertical federated learning model training.

[0093] It is understandable that vertical federated learning can use a variety of encryption methods to ensure the data privacy and security of participants. The following uses homomorphic encryption as an example to illustrate the training process of the vertical federated learning model.

[0094] 3.2. Model training for vertical federated learning:

[0095] For example, taking the slave participant as VFL client A, the master participant as VFL client B, and the trusted coordinator as VFL server as an example, the vertical federated learning model training process can refer to the process shown in Figure 3.

[0096] Figure 3 is a schematic diagram of a vertical federated learning model training process provided by an embodiment of the present application. As shown in Figure 3, the model training process of vertical federated learning may include the following steps:

[0097] S301, initialization. The initialization includes VFL client A initializing local parameters Θ A (or called the model parameters of VFL client A), VFL client B initializes local parameters Θ B (or called the model parameter of VFL client B), the VFL server creates a homomorphic encryption key pair and sends the public key to VFL client A and VFL client B.

[0098] It is understood that the VFL server can also use encryption methods other than homomorphic encryption to ensure the data privacy and security of the participants, and there is no limitation on this.

[0099] S302: VFL client A calculates local parameters as well as in, The symbol represents homomorphic encryption. It should be understood that the result of adding (multiplying) the homomorphic encrypted ciphertext is equal to the ciphertext of the result of adding (multiplying) the plaintext, for example

[0100] It can be calculated based on the local parameter Θ A The common samples are determined by the data of the common samples. The common samples are the common samples between the dataset of VFL client A used for longitudinal federated learning model training and the dataset of VFL client B used for longitudinal federated learning model training, that is, the same sample space between the two datasets. It can be understood that, as shown in Figure 1, longitudinal federated learning first selects common samples from the datasets of multiple participants, that is, the samples in Figure 1: users 1 to 4. In this way, participant A can use the sample data corresponding to users 1 to 4 in participant A's dataset for longitudinal federated learning model training, and participant B can use the sample data corresponding to users 1 to 4 in participant B's dataset for longitudinal federated learning model training.

[0101] Assume that VFL client A provides dataset D for longitudinal federated learning model trainingA Dataset D for longitudinal federated learning model training provided by VFL client B B The common sample identifiers between them may include: n sample identifiers, so that the samples used by VFL client A for longitudinal federated learning model training with VFL client B include samples from D A The n samples corresponding to the n sample identifiers determined in the VFL client A are sampled according to the sample data of the n samples. Sure

[0102] For example, and Θ A The relationship between can be determined according to formula (1).

[0103] in, express The data of the i-th sample in .

[0104] It is understood that formula (1) is only an example, and other methods can also be used according to Θ A and Sure For example Equal to Θ A Transpose or conjugate transpose and The product between them is not specifically limited in this embodiment of the present application.

[0105] L A It can represent the loss function of VFL client A. λ represents the regularization parameter.

[0106] S303: VFL client A sends a and Correspondingly, VFL client B receives the and

[0107] S304: VFL client B calculates local parameters Intermediate parameters And the loss function

[0108] in, Calculation and similar, It can be calculated based on the local parameter ΘB and VFL client B for the common sample data used for longitudinal federated learning model training (i.e., from D B The data of n samples determined in ).For example,

[0109] Parameter d i According to the parameters And the data label y provided by VFL client B i Determine, for example

[0110] Loss Function The L in L can be A , L B , and L AB OK. Among them,

[0111] S305: VFL client B sends a Accordingly, the VFL server receives the

[0112] S306, VFL server uses private key to decrypt Obtain the decrypted loss function and determine whether the model training iteration process is terminated based on the loss function.

[0113] S307: VFL client B sends a Correspondingly, VFL client A receives from VFL client B

[0114] S308. VFL client A calculates local gradients And add random noise from VFL client A get in, According to d i 、Θ A , and the regularization parameter λ is determined, for example

[0115] It should be understood that VFL client A random noise It can also be called the random mask of VFL client A.

[0116] S309, VFL client B calculates local gradient And add random noise from VFL client B get in, and The calculation method is similar to According to d i 、Θ B , and the regularization parameter λ is determined, for example

[0117] It should be understood that the random noise of VFL client B Also called the random mask of VFL client B.

[0118] S310: VFL client A sends a Accordingly, the VFL server receives the

[0119] S311, VFL client B sends to VFL server Accordingly, the VFL server receives the

[0120] S312, VFL server uses private key to decrypt and Get the decrypted result and

[0121] S313: The VFL server sends a Accordingly, VFL client A receives the

[0122] S314: VFL server sends a message to VFL client B. Accordingly, VFL client B receives the

[0123] S315, VFL client A based on decrypted gradient Update local parameters Where η represents the learning rate.

[0124] S316, VFL client B based on decrypted gradient Update local parameters Where η represents the learning rate.

[0125] As shown in Figure 3, the intermediate calculation results are exchanged during the model training process. Therefore, during the entire model training process, the participants (such as VFL client A and VFL client B) do not know the data and features of the other party. After the model training is completed, the participants only obtain the model parameters on their side, protecting the local data privacy of the participants.

[0126] In addition, the vertical federated learning model training shown in Figure 3 is a two-party vertical federated learning model training, that is, the participants in the vertical federated learning include a master participant and a slave participant. It is understood that vertical federated learning can also be extended to scenarios with more participants (or multi-party vertical federated learning), for example, the participants in vertical federated learning can include a master participant and at least two slave participants.

[0127] It should be understood that the multi-party vertical federated learning model training is similar to the two-party vertical federated learning model training shown in Figure 3, and will not be repeated here.

[0128] It should also be understood that the two-party vertical federated learning model training algorithm shown in Figure 3 is only an example. The model training algorithm for two-party or multi-party vertical federated learning can also adopt other algorithms, such as multi-party multi-classification vertical federated learning based on privacy-preserving label sharing, or multi-party vertical federated learning based on secret sharing, etc. The embodiments of this application do not make specific limitations on this.

[0129] It can be understood that the vertical federated learning reasoning is completed after the vertical federated learning model training is completed (that is, the participants have local parameters), through the collaboration between the master participant and the slave participant. For example, in Figure 3, the master participant VFL client B receives a business prediction request, and VFL client B can send the sample identifier to be predicted (used to indicate the sample to be predicted) and the model type identifier to the slave participant VFL client A, so that the slave participant VFL client A can predict the sample identifier to be predicted, the model type identifier, and the local parameter Θ. A , and local data x for inference A , get the local inference result u of VFL client A A , and then VFL client A can send the local inference result u to VFL client B A , so VFL client B can use u AAnd the local inference result u of VFL client B B , determine the inference result u A +u B ,u A +u B Output inference result of vertical federated learning inference.

[0130] It should be understood that, as shown in Figure 1, vertical federated learning requires that the participants have the same sample space, that is, common samples. Before the vertical federated learning model training begins, it is necessary to first select multiple participants in the vertical federated learning (including master participants and slave participants) and the common sample identifiers between the datasets of the multiple participants for vertical federated learning model training. Then, in step S302 shown in Figure 3, VFL client A can determine the sample data of n common samples. To determine the parameters And in step S304 shown in FIG3 , VFL client B can determine the sample data of n common samples To determine the parameters

[0131] However, the data sets of each participant are different, that is, the participants do not naturally have the same sample space. Therefore, before the vertical federated learning model training begins, an important step is sample alignment, which is described in detail below.

[0132] 3.3. Sample alignment:

[0133] The purpose of sample alignment is to identify common samples (or common sample identifiers) between the datasets of various participants (e.g., master and slave) used for training the longitudinal federated learning model. Taking data privacy into consideration, sample alignment also requires the following conditions to be met:

[0134] Condition 1: The participating parties shall not disclose any sample identifiers other than the common sample identifier.

[0135] Condition 2: All participants cannot disclose any sample information to untrusted third parties to avoid attacks by malicious third parties.

[0136] In other words, each participant cannot perform sample alignment by directly exchanging the sample identifiers of their respective datasets used for training the longitudinal federated learning model. Instead, they encrypt the sample identifiers of the datasets through a sample alignment algorithm and then perform sample alignment.

[0137] In addition, sample alignment can be divided into sample alignment without third-party participation and sample alignment with third-party participation. Sample alignment without third-party participation refers to sample alignment performed between participants based on a sample alignment algorithm, without the participation of a third party independent of the participants. Sample alignment with third-party participation refers to sample alignment achieved by introducing a trusted third party independent of the participants.

[0138] For sample alignment without third-party participation, a commonly used technique is private set intersection (PSI). It is understood that PSI can be divided into two-party sample alignment algorithms and multi-party sample alignment algorithms based on the number of participants in vertical federated learning. Two-party sample alignment algorithms can include the following: PSI based on Diffie-Hellman, PSI based on homomorphic encryption, PSI based on Rivest-Shamir-Adleman (RSA) and hash algorithms, PSI based on RSA blind signatures, or PSI based on oblivious transfer pseudo-random function (OT-PRF).

[0139] The multi-party sample alignment algorithm may include: Freedman multi-party secure intersection protocol, or oblivious programmable pseudo-random function (OPPRF).

[0140] It can be understood that the above method for third-party-free sample alignment is only an exemplary introduction, and other methods or technologies can also be used to achieve third-party-free sample alignment, which is not specifically limited in the embodiments of the present application.

[0141] Figure 4 is a schematic diagram of a two-party sample alignment provided by an embodiment of the present application. As shown in Figure 4, the dataset provided by the main participant for longitudinal federated learning model training is dataset 1 in Figure 4, and the dataset provided by the slave participant for longitudinal federated learning model training is dataset 2 in Figure 4. Among them, dataset 1 includes 3 samples, the sample identifiers of the 3 samples are U1, U2, and U3 respectively, and the feature space of the 3 samples includes feature X3, and the label is Y. Dataset 2 also includes 3 samples, the sample identifiers of the 3 samples are U1, U2, and U4 respectively, and the feature space of the 3 samples includes feature X4 and feature X5.

[0142] The master and slave parties can use a two-party sample alignment algorithm (such as PSI based on Diffie-Hellman, PSI based on homomorphic encryption, or PSI based on RSA blind signature, etc.) to perform the sample alignment process. By exchanging encrypted information between the master and slave parties, the common sample identifiers between data sets 1 and 2 can be determined as: sample identifier U1 and sample identifier U2.

[0143] In addition, as shown in FIG4 , after the slave party determines the common sample identifier (i.e., U1 and U2), it can determine a data set 3 for longitudinal federated learning model training with the master party. The data set 3 includes the data of the sample corresponding to the sample identifier U1 selected from the data set 2, and the data of the sample corresponding to the sample identifier U2. The slave party can then perform step S302 in FIG3 based on the data set 3. Similarly, the master party can determine a data set 4 for longitudinal federated learning model training with the slave party based on the common sample identifier. The data set 4 includes the data of the sample corresponding to the sample identifier U1 selected from the data set 1, and the data of the sample corresponding to the sample identifier U2. The master party can then perform step S304 in FIG3 based on the data set 4.

[0144] It should be understood that Figure 4 above is only an illustrative example of the sample alignment process without the participation of a third party. In the actual sample alignment process, for different sample alignment algorithms, the computing power (or computing amount) requirements of each participant in the sample alignment may be the same or different. The following are several examples to illustrate the computing power requirements between the various participants in the sample alignment process without the participation of a third party.

[0145] Example 1:

[0146] For the sample alignment algorithm of PSI based on Diffie-Hellman, participants A and B agree on a large prime number p. The sample alignment algorithm requires the same computing power for participants A and B.

[0147] For example, the sample alignment process of PSI based on Diffie-Hellman includes the following steps:

[0148] Step ①: Participant A hashes the sample identifier x1 to obtain H(x1), and requires that H(x1) is the primitive root of p.

[0149] Step ②: Party B hashes the sample y1 to obtain H(y1), and requires that H(y1) is the primitive root of p.

[0150] Step 3: Participant A randomly generates an integer a (a is kept secret) and calculates Ka = (H(x1)) aMod p, send Ka to participant B, mod means modular division or remainder.

[0151] Step 4: Party B randomly generates an integer b (b is kept secret) and calculates Kb = (H(y1)) b mod p, send Kb to participant A.

[0152] Step 5: After receiving Kb, Party A calculates Kba = (Kb) a mod p, send Kba to participant B.

[0153] Step 6: After receiving Ka, participant B calculates Kab = (Ka) b mod p, send Kab to participant A.

[0154] It can be understood that if x1=y1, then according to the principle of the Diffie-Hellman algorithm, Kba=Kab, so participant A or participant B can determine whether x1 and y1 are the same by comparing whether Kba and Kab are the same.

[0155] Furthermore, when participant A has sample identifiers: x1, x2, …, x n , participant B has sample identifiers: y1,y2,…,y m , Participant A and Participant B exchange encrypted values ​​based on the above-mentioned sample alignment algorithm, and then Participant A calculates [Kba1, Kba2, …, Kbam] and receives [Kab1, Kab2, …, Kabn] from Participant B. In this way, Participant A takes the intersection of these two sets of data, that is, obtains the common sample identifier between Participant A and Participant B. Participant B can also obtain the common sample identifier in the same way, that is, Participant B calculates [Kab1, Kab2, …, Kabn] and receives [Kba1, Kba2, …, Kbam] from Participant A, and then takes the intersection of the two sets of data.

[0156] It can be understood that according to the relevant description of the sample alignment process of PSI based on Diffie-Hellman, the computational complexity between participants A and B is the same, that is, the sample alignment algorithm based on PSI based on Diffie-Hellman requires the same computing power for the two participants.

[0157] Example 2:

[0158] For PSI based on RSA and hash algorithms, one of participant A and participant B can generate a public key (n, e) and a private key (n, d) and send the public key to the other party. This method has different computing power requirements between participant A and participant B.

[0159] For example, the sample alignment process of PSI based on RSA and hash algorithms includes the following steps:

[0160] Step 1. Party A uses the RSA algorithm to generate a public key (n, e) and a private key (n, d), and sends the public key (n, e) to Party B.

[0161] Step 2: Participant B determines the sample identification set y based on the public key (n, e) B =[y1,y2,…,y i ,…,y k ] and sends Y to participant A.

[0162] in, y i ∈y B .y B is the sample identification set used by participant B for sample alignment. i Generated by participant A and y i The corresponding random number, 1<r i <n, and r i Coprime with n. Participant B uses the public key (n, e) to i Obtained after encryption. i ) is y i Substitute into the hash function H(·) to get.

[0163] Step 3: Participant A performs a preliminary calculation on Y based on the private key (n, d) to obtain Y′.

[0164] Step 4: Participant A determines the sample identification set x based on the private key (n, d) A =[x1,x2,…,x i ,…,x l ]The corresponding parameter X.

[0165] in,

[0166] Understandable, x A is the sample identification set of participant A for sample alignment, X is x A Encrypted information after RSA encryption and hashing.

[0167] Step 5: Party A sends Y′ and X to Party B.

[0168] It is understandable that step 4 can also be performed before step 3. For example, step 4 can be performed simultaneously with step 1, and participant A can send X and public key (n, e) to participant B at the same time. This embodiment of the present application does not make specific limitations on this.

[0169] Step 6: Party B calculates the value of Y′ and r i Determine parameter D B , and D B Take the intersection with X to obtain the encrypted common sample identification set I.

[0170] in,

[0171] Understandable, D B Y′ is unblinded (i.e., removing r i ) after the encrypted information, and then D B and X are both encrypted information after RSA encryption and hash processing, so D B The intersection of and X can be taken to obtain the encrypted common sample identification set I. Participant B can determine the plaintext common sample identification set I' based on the encrypted common sample identification set I.

[0172] Step 7: Party B sends the encrypted common sample identification set I to Party A.

[0173] It can be understood that after receiving the encrypted common sample identification set I, participant A can determine the plaintext common sample identification set I'.

[0174] It can also be understood that, according to the above description of the sample alignment process of PSI based on RSA and hash algorithm, the computational complexity of participant B is greater than that of participant A, that is, the sample alignment algorithm of PSI based on RSA and hash algorithm has different computing power requirements for the two participants.

[0175] It should be understood that for PSI based on homomorphic encryption, similar to PSI based on RSA and hash algorithms, the method has different computing power requirements between the two parties.

[0176] In addition, for the multi-party sample alignment algorithm, similar to the two-party sample alignment algorithm, the computing power requirements between the participating parties may be the same or different, depending on the actual multi-party sample alignment algorithm used.

[0177] However, sample alignment has the following problems:

[0178] Question 1: For sample alignment without third-party participation, since sample alignment requires hash encryption, homomorphic encryption, or RSA encryption of all sample identifiers of the dataset used for vertical federated learning, the computing power requirements of the participants are relatively high, and the sample space size (i.e., the number of sample identifiers) may vary between participants. In this case, different sample alignment algorithms may require different computational requirements for each participant, which may result in one or more participants being unable to support the computing power required for sample alignment, and thus unable to achieve sample alignment, or unable to effectively perform sample alignment. For example, for sample alignment algorithms with different computing power requirements for each participant, if the sample alignment requires a large amount of computation, the participant with the heavier load may not be able to support the completion of the sample alignment process.

[0179] Question 2: For sample alignment involving a third party, when a trusted third party independent of the participants in vertical federated learning participates in the sample alignment, there is currently no implementation plan for how to achieve sample alignment.

[0180] In response to the above technical problems, the embodiments of the present application propose the following technical solutions, which will be described below in conjunction with the accompanying drawings.

[0181] The technical solutions of the embodiments of the present application can be applied to various communication systems, such as wireless network systems, vehicle-to-everything (V2X) communication systems, device-to-device (D2D) communication systems, Internet of Vehicles communication systems, 4G mobile communication systems such as long-term evolution (LTE) systems, world-wide interoperability for microwave access (WiMAX) communication systems, 5G mobile communication systems such as new radio (NR) systems, and future communication systems.

[0182] In the embodiment of the present application, "indication" may include direct indication and indirect indication, and may also include explicit indication and implicit indication. The information indicated by a certain information (such as the first indication information, the second indication information, or the third indication information below) is called information to be indicated. In the specific implementation process, there are many ways to indicate the information to be indicated, such as but not limited to, the information to be indicated can be directly indicated, such as the information to be indicated itself or the index of the information to be indicated. The information to be indicated can also be indirectly indicated by indicating other information, where there is an association between the other information and the information to be indicated. It is also possible to indicate only a part of the information to be indicated, while the other parts of the information to be indicated are known or agreed in advance. For example, the indication of specific information can be achieved by means of the arrangement order of each piece of information agreed in advance (such as specified in the protocol), thereby reducing the indication overhead to a certain extent. At the same time, the common parts of each piece of information can be identified and indicated uniformly to reduce the indication overhead caused by indicating the same information separately.

[0183] In addition, the specific indication method can also be various existing indication methods, such as but not limited to the above-mentioned indication methods and various combinations thereof. The specific details of the various indication methods can be referred to the prior art and will not be repeated herein. As can be seen from the above, for example, when it is necessary to indicate multiple information of the same type, there may be a situation where the indication methods for different information are different. In the specific implementation process, the required indication method can be selected according to specific needs. The embodiment of the present application does not limit the selected indication method. In this way, the indication method involved in the embodiment of the present application should be understood to cover various methods that can enable the party to be indicated to obtain the information to be indicated.

[0184] "Pre-definition" or "pre-configuration" can be implemented by pre-saving corresponding codes, tables or other methods that can be used to indicate relevant information in the device, and the embodiments of the present application do not limit the specific implementation method. Among them, "saving" can mean saving in one or more memories. The one or more memories can be set separately or integrated in an encoder or decoder, a processor, or a communication device. The one or more memories can also be partially set separately and partially integrated in a decoder, a processor, or a communication device. The type of memory can be any form of storage medium, and the embodiments of the present application do not limit this.

[0185] The "protocol" involved in the embodiments of the present application may refer to a protocol family in the communication field, a standard protocol with a similar protocol family frame structure, or a related protocol used in future communication systems. The embodiments of the present application do not make specific limitations on this.

[0186] In the embodiments of the present application, descriptions such as "when...", "in the case of...", "if" and "if" all mean that the device will perform corresponding processing under certain objective circumstances. It does not limit the time, nor does it require the device to perform judgment actions when implemented, nor does it mean that there are other limitations.

[0187] In the description of the embodiments of the present application, unless otherwise specified, " / " indicates that the objects associated with each other are in an "or" relationship. For example, A / B can represent A or B. "And / or" in the embodiments of the present application is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, in the description of the embodiments of the present application, unless otherwise specified, "multiple" refers to two or more than two. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple. In addition, in order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish between identical or similar items with basically the same functions and effects. Those skilled in the art will understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit differences. At the same time, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or design. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way for easy understanding.

[0188] The network architecture and business scenarios described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Ordinary technicians in this field will know that with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0189] To facilitate understanding of the embodiments of the present application, a communication system applicable to the embodiments of the present application is first described in detail using the communication system shown in Figure 5 as an example. For example, Figure 5 is a schematic diagram of the architecture of a communication system applicable to the communication method provided in the embodiments of the present application.

[0190] Figure 5 is a schematic diagram of the architecture of a communication system provided in an embodiment of the present application. As shown in Figure 5, the communication system may include one or more terminal devices, access network devices, and core network devices. Among them, the core network device may deploy network function network elements, and the network function network elements may include, for example, network data analysis function network elements (or NWDAF network elements), application function network elements (or AF network elements), access and mobility management function (AMF) network elements, session management function (SMF) network elements, policy control function (PCF) network elements, network storage function (NRF) network elements, or network exposure function (NEF) network elements, etc., without limitation.

[0191] In FIG5 , the terminal device may be located within the beam / cell coverage of the access network device, and the access network device may provide communication services for the terminal device.

[0192] The terminal device in Figure 5 can be a device with wireless transceiver capabilities or a chip or chip system that can be installed in the device, which can allow users to access the network and is used to provide voice and / or data connectivity to users. The terminal device can also be called a UE, a subscriber unit, a terminal, a mobile station (MS), or a mobile terminal (MT).

[0193] Exemplarily, the terminal device in FIG5 may be a mobile phone, a tablet computer, or a computer with wireless transceiver function. The terminal device may also be a user station, a mobile station, a remote station, a remote terminal device, a mobile terminal device, a user terminal device, a wireless communication device, a user agent, a user device, a cellular phone, a cordless phone, a session initiation protocol (SIP) phone, a wireless local loop (WLL) station, a personal digital assistant (PDA), a handheld device with wireless communication capabilities, a computing device, a processing device connected to a wireless modem, an in-vehicle device, a wearable device, a terminal device in the Internet of Things, a home appliance, a virtual reality (VR) terminal, an augmented reality (AR) terminal, a wireless terminal in industrial control, a wireless terminal in unmanned driving, a wireless terminal in telemedicine, a wireless terminal in a smart grid, a wireless terminal in a smart city, a wireless terminal in a smart home, a vehicle with vehicle-to-vehicle (V2V) communication capabilities, an intelligent connected vehicle, an unmanned aerial vehicle to an unmanned aerial vehicle (UAV) There are no restrictions on drones with U2U (Universal to Vehicle) communication capabilities, terminal devices in future networks, or terminal devices in future evolved public land mobile networks (PLMN).

[0194] The access network device in Figure 5 can be any device deployed in the access network that can communicate wirelessly with a terminal device. It can also be a chip or chip system that can be installed in the above-mentioned device. It can also be a logical node or logical module or a function implemented in software. It can be used to implement functions such as wireless physical control functions, resource scheduling and wireless resource management, wireless access control, and mobility management. Specifically, the network device can be a device that supports wired access or a device that supports wireless access.

[0195] Exemplarily, the access network device may be composed of one or more access network (AN) / radio access network (RAN) nodes. The AN / RAN node may be: a gNB, a transmission reception point (TRP), an evolved NodeB (eNB), a radio network controller (RNC), a NodeB (NB), a base station controller (BSC), a base transceiver station (BTS), a home base station (e.g., a home evolved NodeB, or a home NodeB, HNB), a base band unit (BBU), or a wireless fidelity (Wi-Fi) access point (AP).

[0196] In another example, the access network device may also be a device including a centralized unit (CU) node, or a distributed unit (DU) node, or a CU node and a DU node. For example, the access network device can be divided into CU and DU from a logical function perspective, with some protocol layer functions placed under central control in the CU, and the remaining part or all of the protocol layer functions distributed in the DU, which is centrally controlled by the CU. Furthermore, the centralized unit CU can also be divided into a control plane (CU-CP) and a user plane (CU-UP). In different systems, CU (including CU-CP or CU-UP) or DU may also have different names. For example, in an open radio access network (O-RAN) system, CU may also be called O-CU (open CU), DU may also be called O-DU, CU-CP may also be called O-CU-CP, and CU-UP may also be called O-CU-UP.

[0197] The network data analysis function network element in Figure 5 has functions such as data collection, model training, data analysis, and model reasoning. It can be used to collect relevant data from network function network elements, third-party service servers, terminal devices, or network management systems (such as OAM), perform data analysis or model training based on the relevant data, and provide data analysis results to network function network elements, third-party service servers, terminal devices, or network management systems, or provide trained models to other data analysis function network elements. Network data analysis function network elements can be divided into analysis logic functions and model training logic functions based on their functions. The analysis logic function is the logic function within the network data analysis function network element, which is used to perform model reasoning, derive analysis results (i.e., derive statistical or predictive analysis results based on the analysis consumer's request), and make the analysis results available. The model training logic function is the logic function within the network data analysis function network element, which is used to train models and make training services available (for example, providing trained models). A network data analysis function network element may include only analysis logic functions, only model training logic functions, or both analysis logic functions and model training logic functions. This is not specifically limited in the embodiments of the present application.

[0198] The application function network element in Figure 5 is mainly an intermediate functional entity that provides interaction between application servers in the data network (DN) and network elements in the core network, and transmits the requirements of the application side to the network side (for example, service quality requirements or user status event subscription, etc.). The application server can use it to dynamically control the network service quality and billing, obtain the operating information of a certain network element in the core network, etc. In the embodiment of the present application, the application function network element can be a functional entity deployed by the operator or a functional entity deployed by the service provider. The service provider can be a third-party service provider or a service provider within the operator, without limitation.

[0199] Among them, the access and mobility management functional network element in Figure 5 is mainly responsible for access authentication of terminal devices, mobility management, signaling interaction between various functional network elements, and termination of non-access stratum (NAS) layer signaling security. For example, it manages the user's registration status, reachability status, N1 / N2 interface signaling transmission, access authentication and authorization, user connection status, user registration and network entry, tracking area update, cell switching user authentication, key security, etc.

[0200] Among them, the session management function network element in Figure 5 mainly provides session management of terminal device sessions (such as session establishment, modification, and release), network protocol (Internet protocol, IP) address allocation and management, and user plane network element selection and control.

[0201] Among them, the policy control function network element in Figure 5 is mainly responsible for generating policies such as terminal device access policy and service quality flow control policy, and can also provide the generated policies to the access and mobility management function network element or the session management function network element.

[0202] Among them, the network storage function network element in Figure 5 can provide storage capabilities for contract data, policy data and capability exposure related data.

[0203] Among them, the network open function network element in Figure 5 is mainly responsible for providing network capabilities and event opening to external access entities (such as AF network elements), and receiving relevant external information (such as receiving information provided by AF network elements).

[0204] The following describes the participants in vertical federated learning with reference to Figure 5.

[0205] The master or slave participant in vertical federated learning can be a network function network element such as the aforementioned network data analysis function network element, application function network element, access and mobility management function network element, session management function network element, or policy control function network element. It is understood that the master or slave participant in vertical federated learning can also be a terminal device or an access network device, etc., and this embodiment of the application does not specifically limit this.

[0206] Among them, the trusted coordinator among the participants of the vertical federated learning shown in Figure 5 can be a functional network element independently deployed in the core network, or can be deployed in the network function network element in Figure 5 (for example, a network storage function network element, or a network data analysis function network element, etc.). It can be understood that the trusted coordinator can also be called a vertical federated learning server (VFL server), which is not specifically limited in this embodiment of the present application.

[0207] Optionally, the communication system shown in FIG5 further includes a vertical federated learning support function (VFL support function, VFLSF) network element, which is responsible for the registration and discovery of participants in vertical federated learning. It should be understood that the VFLSF network element can be deployed independently or on the above-mentioned network data analysis function network element, application function network element, network storage function network element, or other network elements (such as network open function network element).

[0208] In addition, the trusted coordinator can be deployed in the same network element as the VFLSF network element, or can be deployed separately, which is not specifically limited in the embodiments of the present application.

[0209] It should be noted that the terminal device, access network device, and core network device in the embodiments of the present application can be one or more chips, or a system on chip (SOC). Figure 5 is only an exemplary figure, and the number of devices or equipment included is not limited. The names of the various devices or equipment and the names of the various links in Figure 5 are not limited. In addition to the names shown in Figure 5, the various devices or equipment and the various links can also be named other names without limitation.

[0210] In addition, in addition to the device or network element shown in Figure 5, the communication system may also include other devices or network elements, such as the OAM, user plane function (UPF) network element, or application server shown in Figure 5, without limitation.

[0211] It should be understood that as the network evolves, the communication system shown in FIG5 may also support or include other network functions, such as AI or perception-related functions. In addition, the network element names shown in FIG5 may also change in future evolving networks, and this embodiment of the present application does not specifically limit this.

[0212] In response to question 1, an embodiment of the present application provides an information transmission method, the execution subject of the method can be a first network element, the first network element can be a participant in the vertical federated learning model training (for example, a master participant or a slave participant), the first network element can be a network function network element such as the network data analysis function network element, application function network element, access and mobility management function network element, session management function network element, or policy control function network element in Figure 5, or it can be a terminal device or access network device, etc., and the embodiment of the present application does not specifically limit this. It should be understood that the second network element can also be a participant in the vertical federated learning model training.

[0213] In one possible implementation, a first network element sends a first request to a second network element and receives a first response from the second network element. The first request includes first information indicating at least one first sample alignment algorithm supported by the first network element. The first response indicates whether the second network element supports a second sample alignment algorithm determined from the at least one first sample alignment algorithm, or does not support sample alignment. The second sample alignment algorithm is used to determine common samples between a dataset used for training a longitudinal federated learning model of the first network element and a dataset used for training a longitudinal federated learning model of the second network element. In this way, by sending the first request to the second network element, the first network element can cause the second network element to determine, based on the at least one first sample alignment algorithm, whether the second network element and the first network element support the same sample alignment algorithm for sample alignment. If it is determined that the second network element and the first network element support the same sample alignment algorithm, the second network element can feedback the second sample alignment algorithm to the first network element via the first response. The first and second network elements can then negotiate the sample alignment algorithm to be used in the sample alignment process, thereby improving the efficiency and success rate of sample alignment between the first and second network elements. Furthermore, when the second network element does not support at least one first sample alignment algorithm, the second network element can provide feedback to the first network element through a first response indicating that it does not support sample alignment, so that the first network element can promptly perform sample alignment with other participants in the vertical federated learning model training, thereby improving the efficiency of the first network element in performing sample alignment. Therefore, based on the information transmission method provided in the embodiment of the present application, the efficiency and success rate of the first network element in performing sample alignment can be improved.

[0214] In response to question 2, an embodiment of the present application provides an information transmission method, the execution subject of which may be a first network element, which may be a trusted third party, such as the trusted coordinator in Figure 5, or a VFLSF network element, etc.; or, the first network element may be a participant in the training of a vertical federated learning model (such as a master participant or a slave participant).

[0215] In one possible implementation, a first network element sends a first request to a second network element and receives a first response from the second network element. The first request includes first indication information, and the first indication information is used to indicate the provision of a sample identifier for longitudinal federated learning model training, and the first response includes the first sample identifier. In this way, by sending the first request to the second network element, the first network element can enable the second network element to feedback the first sample identifier to the first network element, and then the first network element can determine the common sample identifiers between the various participants participating in the longitudinal federated learning model training based on the first sample identifier, thereby achieving sample alignment. For example, the first network element can be a VFL server, and the second network element can be a master participant or a slave participant in the longitudinal federated learning. Then, the first network element can obtain the sample identifier corresponding to the data set of the second network element for longitudinal federated learning model training by sending the first request to the second network element, so that the first network element can determine the common samples between the various participants participating in the same longitudinal federated learning model training, thereby achieving sample alignment. For another example, the second network element may be a VFLSF network element, and the first network element may be a master participant, a slave participant, or a coordinator in the vertical federated learning model training. Then, the second network element may respond to the first request in the first network element and feedback to the first network element the common sample identifiers of at least two network elements participating in the same vertical federated learning model training, thereby achieving sample alignment.

[0216] For ease of understanding, the following is a unified introduction to the technical terms involved in the embodiments of this application:

[0217] 1) Samples: The samples in the embodiments of this application refer to the samples contained in the dataset used for training the longitudinal federated learning model. Samples can be, for example, users, accounts, applications, terminal devices, sensory targets (such as vehicles or transportation infrastructure), or images. Samples can be indicated or characterized by sample identifiers. For details, see Figure 1 and Table 2.

[0218] 2) Common samples: Common samples refer to the common samples between the datasets of all participants used for training the longitudinal federated learning model. For example, as shown in Figure 1, the common samples are users 1 to 4.

[0219] In addition, the common sample can be indicated by a common sample identifier. For example, as shown in FIG4 , the common sample identifiers are sample identifiers U1 and U2, and the common sample is the sample indicated by the sample identifier U1 and the sample indicated by the sample identifier U2.

[0220] 3) Sample alignment: The purpose of sample alignment is to determine the common samples (or common sample identifiers) between the data sets of each participant for longitudinal federated learning model training. At the same time, other sample identifiers other than the common sample identifier cannot be leaked between the participants, and each participant cannot leak any sample information to an untrusted third party.

[0221] It can be understood that when each participant has determined a common sample, each participant can determine the sample used for longitudinal federated learning model training, and then conduct longitudinal federated learning model training.

[0222] It can also be understood that the common sample identifier or the name of the common sample is only an example. The common sample identifier can also be called an intersection sample ID or a public sample identifier, and the common sample can be called an intersection sample or a public sample, etc. The embodiments of the present application do not make specific limitations on this.

[0223] 4) Sample Alignment Type: Sample alignment can be divided into two-party sample alignment and multi-party sample alignment. Two-party sample alignment refers to a vertical federated learning model training involving a master and a slave. Multi-party sample alignment refers to a vertical federated learning model training involving a master and at least two slaves. Participants in the vertical federated learning model training can support two-party sample alignment and / or multi-party sample alignment.

[0224] It can be understood that whether a participant supports two-party sample alignment and / or multi-party sample alignment depends on the sample alignment algorithm supported by the participant. If participant A supports programmable oblivious pseudo-random function and PSI based on homomorphic encryption, then participant A supports two-party sample alignment and multi-party sample alignment; if participant A supports two-party sample alignment algorithms such as PSI based on Diffie-Hellman or PSI based on homomorphic encryption, then participant A only supports two-party sample alignment; if participant A supports Freedman multi-party secure intersection protocol or multi-party sample alignment algorithms such as programmable oblivious pseudo-random function, then participant A only supports multi-party sample alignment.

[0225] 5) Sample Alignment Algorithm: The sample alignment algorithm is used to identify common samples between at least two participants' datasets used for training a longitudinal federated learning model. For example, the sample alignment algorithm can encrypt sample identifiers. The two participants then interact with each other to identify common samples between their datasets. This prevents the leakage of sample identifiers and protects user privacy.

[0226] The following will describe in detail the interaction process between the network elements / devices in the above communication system through a method embodiment in conjunction with Figures 6 to 12. The information transmission method provided in the embodiment of the present application can be applied to the communication system shown in Figure 5 above.

[0227] Figure 6 is a flowchart of an information transmission method provided in an embodiment of the present application. The information transmission method is a sample alignment method applicable to problem 1 without third-party participation, which mainly involves the interaction between the main participant, at least one candidate slave participant, a trusted coordinator (i.e., VFL server), and a vertical federated learning support function (VFLSF) network element. Among them, sample alignment without third-party participation means that there is no third-party participation in the sample alignment process between the main participant and at least one candidate slave participant. The third party may or may not participate in other processes related to vertical federated learning (such as vertical federated learning model training and / or reasoning), and the embodiment of the present application does not make specific limitations on this.

[0228] In the information transmission method process shown in FIG6 , the main participant initiates vertical federated learning, and the VFL server is responsible for discovering and selecting the participants (or vertical federated learning clients (VFL clients)).

[0229] As shown in FIG6 , the process of the information transmission method includes:

[0230] S601: A participant sends a registration request message to a VFLSF network element. Correspondingly, the VFLSF network element receives the registration request message from the participant.

[0231] The registration request message is used to indicate the participant's sample alignment capability. In other words, the participant sends a registration request message to the VFLSF network element to register the participant's sample alignment capability with the VFLSF network element, making it easier for the VFL server or other participants initiating vertical federated learning to discover the participant, thereby improving the efficiency and success rate of discovering participants.

[0232] In addition, the VFLSF network element in step S601 can be the VFLSF network element in Figure 5, or the network storage function network element, or the network open function network element, and the participants in step S601 can be the network data analysis function network element, application function network element, access and mobility management function network element, session management function network element, or policy control function network element in Figure 5, or other network function network elements, or terminal devices, or access network devices, etc. The embodiments of the present application do not make specific limitations on this.

[0233] In a possible implementation, the sample alignment capability includes at least one of the following: a sample alignment type supported by the participant, a supported sample alignment algorithm, or a time period for supported sample alignment.

[0234] Among them, the sample alignment type may include two-party sample alignment and multi-party sample alignment, and the sample alignment type supported by the participant may mean: the participant only supports two-party sample alignment, or the participant only supports multi-party sample alignment, or the participant supports two-party sample alignment and multi-party sample alignment. In addition, the sample type supported by the participant can also be used to indicate that the participant does not support sample alignment. For example, the protocol may stipulate that type a of the sample alignment type supported by the participant indicates that sample alignment is not supported. In other words, the sample alignment type supported by the participant can be used to indicate whether the participant supports sample alignment, or to indicate whether the participant supports two-party sample alignment, or to indicate whether the participant supports multi-party sample alignment, or to indicate whether the participant supports both two-party sample alignment and multi-party sample alignment.

[0235] It is understandable that the participant does not support sample alignment, for example, the participant may not pre-configure the sample alignment algorithm.

[0236] For example, the sample alignment type supported by the participant can be indicated by the parameter a contained in the registration request message, and the several possibilities of the sample alignment type supported by the above-mentioned participant can be represented by different parameters. For example, parameter a1 is used to indicate whether the participant supports sample alignment, parameter a2 is used to indicate whether the participant supports two-party sample alignment, parameter a3 is used to indicate whether the participant supports multi-party sample alignment, parameter a4 is used to indicate whether the participant supports both two-party sample alignment and multi-party sample alignment, etc. In addition, the several possibilities of the sample alignment type supported by the above-mentioned participants can also be indicated by other parameters or a combination of parameters, which is not specifically limited in the embodiments of the present application.

[0237] In addition, the types of sample alignment supported by the participants can be used to determine the types of sample alignment algorithms supported by the participants. For example, if the participants support two-party sample alignment, it can be determined that the sample alignment algorithms supported by the participants include two-party sample alignment algorithms. For another example, if the participants support multi-party sample alignment, it can be determined that the sample alignment algorithms supported by the participants include multi-party sample alignment algorithms.

[0238] The sample alignment algorithms supported by the participants may include at least one sample alignment algorithm. That is, the sample alignment algorithms supported by the participants can be used to determine the specific sample alignment algorithms supported by the participants, facilitating subsequent discovery of the participants by the VFL server. Furthermore, the sample alignment algorithm may be a two-party sample alignment algorithm and / or a multi-party sample alignment algorithm mentioned above, such as PSI based on Diffie-Hellman, PSI based on homomorphic encryption, PSI based on RSA and hash algorithms, or the Freedman Multi-Party Secure Intersection Protocol.

[0239] Exemplarily, the sample alignment algorithms supported by the participants can be represented by the parameter supported sample alignment algorithm (supported sample alignment algorithm), which can be a value list used to represent a group of sample alignment algorithms. For example, the parameter can be used to indicate [PSI based on Diffie-Hellman, PSI based on homomorphic encryption, PSI based on RSA and hash algorithm], and then the parameter can be used to determine whether the participant supports the use of the above three algorithms for sample alignment. For another example, the protocol can be agreed upon, or the participants and the VFLSF network element can be negotiated in advance to use symbols (such as a set of numbers and / or characters) to represent a sample alignment algorithm, such as using "01" to represent PSI based on Diffie-Hellman, using "02" to represent PSI based on homomorphic encryption, and using "03" to represent PSI based on RSA and hash algorithm. If the participant only supports PSI based on Diffie-Hellman and PSI based on RSA and hash algorithm, the parameter can be expressed as [01,03].

[0240] It should be understood that the sample alignment algorithm may also be other sample alignment algorithms besides PSI, and the embodiments of the present application do not specifically limit this.

[0241] The time period that a participant supports for sample alignment can be used to indicate the time period that the participant supports (or expects) for sample alignment. This time period can be determined by the participant based on changes in its own computing power or load. For example, if a participant determines that its load is low during the time periods [9:00-11:00] and [14:00-17:00] and can support sample alignment, the participant can determine the time periods supported for sample alignment as [9:00-11:00] and [14:00-17:00]. In other words, the time period that the participant supports for sample alignment can be used to determine the specific time period that the participant supports for sample alignment, further improving the efficiency of subsequent VFL server discovery of participants.

[0242] It can be understood that the time period in which the participant supports sample alignment can also be indicated by a parameter. For example, the time period in which the participant supports sample alignment can be indicated by the parameter supported sample alignment time period (supported sample alignment time period), which can indicate a start time (for example, 9:00) and an end time (for example, 11:00); or, the parameter can indicate a start time (for example, 9:00 or 14:00) and a duration (for example, 1 hour, or 2 hours, etc.); or, the parameter can also indicate a time offset and a reference time, and the time offset can be a time offset relative to the reference time, and then the two time endpoints corresponding to the time period for performing sample alignment can be determined by the time offset and the reference time. For example, one of the two time endpoints can be the reference time, and the other time endpoint can be the difference or sum between the reference time and the time offset; or, the two time endpoints are the difference between the reference time and the time offset and the sum between the reference time and the time offset, respectively.

[0243] In addition, the time period for sample alignment involved in the embodiments of the present application can be understood as the time period within which the participants can perform sample alignment. For example, the participants can start performing sample alignment at the start time of the above-mentioned time period, or they can start performing sample alignment at other times within the above-mentioned time period except the start time. The embodiments of the present application do not make specific limitations on this.

[0244] It should be understood that the time indicated above may be standard time, which may be Coordinated Universal Time (UTC), Global Positioning System (GPS) time, Long Range Navigator (LORAN) time, or International Atomic Time (TAI), etc., and the embodiments of the present application do not specifically limit this. In addition, standard time may also be referred to as absolute time, which is objective and has nothing to do with any particular reference system. All participants in vertical federated learning and VFLSF network elements have the same understanding of the same absolute time.

[0245] In addition, the above implementation method of indicating the time period for participants to support sample alignment is only an example. Other methods can also be used to indicate the time period for participants to support sample alignment. The embodiments of the present application do not specifically limit this.

[0246] Exemplarily, the sample alignment capability may be indicated by a parameter sample alignment capability indication, which may be used to indicate the sample alignment type supported by the participant, the sample alignment algorithm supported, or the time period for which sample alignment is supported.

[0247] It is understood that the participant may send a registration request message to the VFLSF network element based on the identifier or address of the VFLSF network element. The identifier or address of the VFLSF network element may be pre-configured by the protocol, or received by the participant from another network element. For example, the participant may receive the identifier or address of the VFLSF network element from a network storage function network element. This embodiment of the present application does not specifically limit this.

[0248] For example, the address of the VFLSF network element may be an Internet Protocol (IP) address, or a fully qualified domain name (FQDN), or a uniform resource locator (URL).

[0249] It should be understood that in the embodiment of the present application, the participants can be divided into master participants and slave participants as shown in Figure 2, or the participants in the embodiment of the present application can also be divided into master participants and slave participants, which is not limited to this.

[0250] For ease of understanding, the following explanation is given by taking the distinction between the participants as the main participants and the sub-participants as an example.

[0251] In a possible implementation, the participants in the embodiment of the present application are multiple participants, and the multiple participants can be divided into a master participant and a slave participant. Then, step S601 may include:

[0252] S601a: The primary participant sends a registration request message to the VFLSF network element. Correspondingly, the VFLSF network element receives the registration request message from the primary participant. The registration request message is used to indicate the sample alignment capability of the primary participant.

[0253] S601b: The slave participant sends a registration request message to the VFLSF network element. Correspondingly, the VFLSF network element receives the registration request message from the slave participant. The registration request message is used to indicate the sample alignment capability of the slave participant.

[0254] It can be understood that the main participant can be the application function network element, network data analysis function network element, or other network function network elements (such as the access and mobility management function network element, or session management function network element in Figure 5), and the slave participant can also be the application function network element, network data analysis function network element, or other network function network elements. The embodiments of the present application do not make specific limitations on this.

[0255] S602: The VFLSF network element sends a registration response message to the participant. Correspondingly, the participant receives the registration response message from the VFLSF network element.

[0256] The registration response message is used to indicate whether the registration is successful. For example, the registration response message may include a result indication parameter, which may be used to indicate whether the registration is successful.

[0257] It can be understood that, as described in step S601, there are multiple participants, and the multiple participants can be divided into a master participant and a slave participant. Then, step S602 may include:

[0258] S602a: The VFLSF network element sends a registration response message to the primary participant. Correspondingly, the primary participant receives the registration response message from the VFLSF network element.

[0259] S602b: The VFLSF network element sends a registration response message to the slave party. Correspondingly, the slave party receives a registration response message from the VFLSF network element.

[0260] It can be understood that the above steps S601 and S602 are the sample alignment capability registration process. After the participant's sample alignment capability is successfully registered, it can be prepared for the subsequent discovery and selection of participants to initiate vertical federated learning. In addition, when the participant's sample alignment capability (for example, the sample alignment type supported by the participant, the supported sample alignment algorithm, or the time period supported for sample alignment) changes, the participant can update the participant's sample alignment capability by re-performing the above sample alignment capability registration process.

[0261] It should be understood that the registration request message of the above step S601 can also be used to register other information of the participant in addition to the sample alignment capability, such as the participant's identification, address, the type of vertical federated learning capability supported by the participant, the vertical federated learning group identification (VFL group ID) supported by the participant, or the vertical federated learning model type identification (model type ID(s)) supported by the participant, etc. The embodiment of the present application does not make specific limitations on this.

[0262] It is understood that the type of vertical federated learning capability supported by a participant can be used to indicate whether the participant supports serving as a participant in vertical federated learning, or whether the participant supports serving as a master participant and / or a slave participant in vertical federated learning. Whether a participant supports serving as a participant in vertical federated learning can refer to whether the participant supports serving as a participant in vertical federated learning model training and / or vertical federated learning inference. Furthermore, whether a participant supports serving as a master participant and / or a slave participant in vertical federated learning can refer to whether the participant supports providing data labels and / or data features for federated learning tasks.

[0263] In addition, the VFL server may also send a registration request to the VFLSF network element to request registration of the VFL server information. The VFL server information may include: the VFL server identifier, address, vertical federated learning capability type supported by the VFL server, vertical federated learning group identifier (VFL group ID) supported by the VFL server, or vertical federated learning model type identifier (model type ID(s)) supported by the VFL server, etc.; alternatively, the VFLSF network element may configure the VFL server information in advance, which is not specifically limited in this embodiment of the present application.

[0264] It can be understood that the vertical federated learning capability type supported by the VFL server can be used to indicate whether the VFL server supports serving as the coordinator of vertical federated learning, that is, whether the VFL server supports functions such as maintaining the vertical federated learning process and authorizing the admission and removal of federated learning members.

[0265] In addition, in an embodiment of the present application, the vertical federated learning process may include: a sample alignment process, a vertical federated learning model training process, or a vertical federated learning reasoning process, which are uniformly described here and will not be repeated below.

[0266] It should be understood that in the process shown in Figure 6, the main participant can initiate vertical federated learning. For example, the main participant determines to train the model through vertical federated learning (for example, the main participant finds that the training data cannot be collected due to privacy protection issues), and then initiates vertical federated learning. In addition, in the vertical federated learning process, the VFL server can not only be responsible for decrypting and transmitting the intermediate calculation results in the vertical federated learning model training process, but also be responsible for discovering (or selecting) slave participants. The main participant can discover the VFL server that meets the vertical federated learning requirements initiated by the main participant through the VFLSF network element, and then discover the slave participant through the VFL server to carry out subsequent sample alignment, vertical federated learning model training, or vertical federated learning reasoning and other processes.

[0267] Steps S603 and S604 are described below. Steps S603 and S604 are used by the primary participant to discover a VFL server that meets the requirements of the vertical federated learning initiated by the primary participant.

[0268] S603: The primary participant sends a discovery request message to the VFLSF network element. Correspondingly, the VFLSF network element receives the discovery request message from the primary participant. The discovery request message is used to request the discovery of a VFL server that meets the vertical federated learning requirements initiated by the primary participant.

[0269] In a possible implementation, the discovery request message includes at least one of the following: a vertical federated learning group identifier, a vertical federated learning model type identifier, or a first vertical federated learning capability type indication.

[0270] Among them, the vertical federated learning group identifier (VFL group ID) may refer to the vertical federated learning group that the main participant expects from multiple candidate vertical federated learning groups, and the vertical federated learning group identifier may be used to determine the vertical federated learning group that the main participant expects from multiple candidate vertical federated learning groups. Each candidate vertical federated learning group in the multiple candidate vertical federated learning groups may include multiple slave participants and / or at least one VFL server. It can be understood that the multiple candidate vertical federated learning groups may be pre-configured, or negotiated in advance between the main participant, the slave participant, and the VFLSF network element, or indicated by the VFLSF network element, and the embodiment of the present application does not specifically limit this.

[0271] In addition, the discovery request message may include one or more vertical federated learning group identifiers. The discovery request message includes multiple vertical federated learning group identifiers, that is, the main participant expects to perform vertical federated learning based on multiple vertical federated learning groups.

[0272] The vertical federated learning model type identifier (model type ID(s)) may refer to the type of vertical federated learning model that the primary participant desires to train, such as a business experience model or a personalized recommendation model. It is understood that the type of vertical federated learning model and its corresponding identifier may be pre-configured by the protocol or provided by the VFL server, and this embodiment of the present application does not specifically limit this.

[0273] In addition, there can be multiple vertical federated learning model type identifiers, that is, the main participant can obtain one or more VFL servers that support multiple vertical federated learning model type identifiers.

[0274] The first vertical federated learning capability type indication is used to indicate the discovery of a VFL server. The indication information may be represented by the parameter VFL capability type=VFL server.

[0275] Optionally, the discovery request message may also indicate: the time period for executing longitudinal federated learning and the supported training modes. The time period for executing longitudinal federated learning may include: the time period for executing longitudinal federated learning model training, the time period for executing longitudinal federated learning reasoning, or the time period for executing sample alignment. The supported training mode may refer to a mode or process for longitudinal federated learning model training, which may be a model training process as shown in FIG3 or other model training processes, and the embodiments of the present application do not specifically limit this.

[0276] It can be understood that the desired time period for performing vertical federated learning and the supported training modes can be used by the VFLSF network element to better discover a VFL server that meets the requirements of the primary participant.

[0277] In addition, the main participant can send the actual code corresponding to the training mode; or, the protocol can be agreed upon or the participants and the VFLSF network element can negotiate in advance to use symbols (such as numbers and / or characters) to represent a training mode, for example, symbol c1 can represent training mode 1, symbol c2 can represent training mode 2, symbol c3 can represent training mode 3, etc., or other methods can be used to indicate the training mode, which is not specifically limited in the embodiments of the present application.

[0278] S604: The VFLSF network element sends a discovery response message to the primary participant. Correspondingly, the primary participant receives the discovery response message from the VFLSF network element.

[0279] The discovery response message indicates information about at least one VFL server. The at least one VFL server may be at least one VFL server determined by the VFLSF network element from among multiple VFL servers based on the discovery request message. The information about the at least one VFL server may include at least one VFL server instance, each VFL server instance being represented by a VFL server identifier and / or address.

[0280] It can be understood that after the master participant discovers the VFL server through steps S603 and S604, it can discover the slave participant through the VFL server to facilitate subsequent sample alignment and vertical federated learning model training.

[0281] Steps S605 to S609 are described below. Steps S605 to S607 are used to discover candidate slave participants (or potential slave participants) that can perform sample alignment with the master participant.

[0282] S605: The primary participant sends a vertical federated learning initiation request message to the VFL server. Correspondingly, the VFL server receives the vertical federated learning initiation request message from the primary participant.

[0283] The VFL server is a VFL server determined by the primary participant from at least one VFL server. For example, if the discovery response message indicates multiple VFL servers, the primary participant may randomly select a VFL server from the multiple VFL servers. Alternatively, the primary participant may select a VFL server from the multiple VFL servers using other methods, which are not specifically limited in this embodiment of the present application.

[0284] It can be understood that the vertical federated learning initiation request message is used to initiate the vertical federated learning process, and may also include the meaning of indicating the selection of a slave participant or requesting the selection of a slave participant.

[0285] In one possible implementation, the vertical federated learning initiation request message includes the following parameters: a vertical federated learning group ID (VFL group ID) and a vertical federated learning model type ID (mode type ID(s)). It will be appreciated that the vertical federated learning group ID and the vertical federated learning model type ID(s) can be used to select a slave participant. The specific meanings of the vertical federated learning group ID and the vertical federated learning model type ID can be found in step S603 above and will not be further described.

[0286] Optionally, the vertical federated learning initiation request message also includes the following parameters:

[0287] Master VFL client indicator: The master VFL client indicator is used to indicate that the initiator of the vertical federated learning initiation request message is the master client.

[0288] VFL client selection flag: The VFL client selection flag is used to indicate the selection of a participant or a slave participant.

[0289] Alternatively, the slave VFL client requirement is used to indicate a requirement on a slave VFL client, or to indicate a condition for selecting a slave VFL client.

[0290] It can be understood that vertical federated learning can be initiated by the master participant or the slave participant. In this way, the master participant's indication can be used to determine whether the initiator of the vertical federated learning initiation request message is the master participant or the slave participant, so that the VFL server can determine whether the master participant needs to be selected.

[0291] For example, if the vertical federated learning initiation request message carries a master participant indication, the VFL server determines to select a slave participant; if the vertical federated learning initiation request message carries a slave participant indication to indicate that the initiator of the vertical federated learning initiation request message is a slave participant, the VFL server determines to select the master participant, or the VFL server determines to select the master participant and at least one slave participant.

[0292] The participant selection flag is used to indicate a request to select a participant (i.e., select a VFL client) or to indicate a request to select a slave participant (i.e., select a slave VFL client). If the vertical federated learning initiation request message includes a master participant indication or a slave participant indication, the vertical federated learning initiation request message may not carry the participant selection flag.

[0293] The requirements of the slave party may specifically include: a time interval supporting VFL, and / or a desired training mode. The time interval supporting VFL may include: a time interval supporting VFL model training, a time interval supporting VFL reasoning, or a time interval supporting sample alignment. The desired training mode may refer to a specific training mode or training process that the master party expects to use.

[0294] It should be understood that the time period supporting vertical federated learning can be one or more time periods, and the desired training mode can be one or more training modes. If the time period supporting vertical federated learning is multiple time periods, the VFL server can select a participant based on one of the time periods. Similarly, if the desired training mode is multiple training modes, the VFL server can select a participant based on one of the training modes.

[0295] It is understood that after the VFL server receives the vertical federated learning initiation request message from the master participant, it can determine the conditions for selecting the slave participant based on the parameters carried in the vertical federated learning initiation request message, thereby triggering the participant discovery process (also known as the VFL client discovery process). In addition, the participant discovery process can also be understood as the participant selection process (also known as the VFL client selection process).

[0296] The following introduces steps S606 to S607, which are the participant discovery process.

[0297] S606: The VFL server sends a participant discovery request message to the VFLSF network element. Correspondingly, the VFLSF network element receives the participant discovery request message from the VFL server.

[0298] The participant discovery request message is used to request the discovery of participants. The participant discovery request message may include: a vertical federated learning group identifier, a vertical federated learning model type identifier, slave participant requirements, a second vertical federated learning capability type indication, and a sample alignment requirement.

[0299] It can be understood that the vertical federated learning group identifier and the vertical federated learning model type identifier can be specifically referred to the relevant description of step S603 and will not be repeated here.

[0300] In addition, when the vertical federated learning initiation request message in step S605 includes a slave participant requirement, the slave participant requirement in step S606 is the same as the slave participant requirement in step S605, or the slave participant requirement in step S606 is determined from the slave participant requirement in step S605. This embodiment of the present application does not specifically limit this.

[0301] In the case where the vertical federated learning initiation request message in step S605 does not include the slave participant requirements, the VFL server can determine the slave participant requirements by itself, or the VFL server can negotiate the slave participant requirements with the master participant. This embodiment of the present application does not specifically limit this.

[0302] The second vertical federated learning capability type indication is used to indicate a request to discover a slave participant, and may be represented by, for example, VFL capability type=slave VFL client.

[0303] The sample alignment requirement (or required sample alignment capability) may be used to indicate the sample alignment capability that the requesting party should support. The sample alignment capability may include, for example, a sample alignment type, a sample alignment algorithm, or a time period for sample alignment.

[0304] For example, the sample alignment requirement may indicate that the requesting party should support the sample alignment type, such as indicating that the requesting party should support sample alignment, or indicating that the requesting party should support two-party sample alignment, or indicating that the requesting party should support multi-party sample alignment, or indicating that the requesting party should support both two-party sample alignment and multi-party sample alignment.

[0305] It can be understood that the VFLSF network element can select a participant based on the sample alignment type supported by the participant carried in the registration request message sent by the participant in step S601, and the sample alignment type that the participant indicating the request should support.

[0306] For another example, the sample alignment requirement may indicate the sample alignment algorithm that the requesting party should support. If the sample alignment algorithm indicated as supported by the requesting party only includes one sample alignment algorithm, the discovered party supports the sample alignment algorithm. If the sample alignment algorithm indicated as supported by the requesting party includes multiple sample alignment algorithms, the discovered party only needs to support any one of the multiple sample alignment algorithms, or it may also indicate that the discovered party should support the multiple sample alignment algorithms at the same time.

[0307] It can be understood that the VFLSF network element can select a participant according to the sample alignment algorithm (supported sample alignment algorithm) supported by the participant carried in the registration request message sent by the participant in step S601 and the sample alignment algorithm that the requesting participant should support as indicated in step S606.

[0308] For another example, the sample alignment requirement may indicate an expected sample alignment time period. It is understood that the VFLSF network element may select a participant based on the supported sample alignment time period carried in the registration request message sent by the participant in step S601 and the expected sample alignment time period.

[0309] It can be understood that when the participant finds that the request message carries the sample alignment requirement, the VFLSF network element can select a candidate slave participant (or potential slave participant) that can perform sample alignment with the main participant, thereby improving the success rate and efficiency of negotiating the sample alignment method (such as sample alignment type, algorithm, and time period for performing sample alignment) between the main participant and the candidate slave participant.

[0310] It should be understood that for the sample alignment capability, please refer to the relevant description of the sample alignment capability in step S601, which will not be repeated here.

[0311] In addition, the VFL server can pre-configure multiple sample alignment types and multiple sample alignment algorithms so that the VFL server can indicate the sample alignment capabilities that the requesting party should support. It is understood that multiple sample alignment types and multiple sample alignment algorithms can also be negotiated in advance between the VFL server and the VFLSF network element, or indicated by the VFLSF network element, and this embodiment of the application does not specifically limit this.

[0312] It should be understood that the vertical federated learning initiation request message in step S605 may also include a sample alignment requirement, and the sample alignment requirement included in the participant discovery request message in step S606 is one or more of the sample alignment requirements included in the vertical federated learning initiation request message in step S605. If the vertical federated learning initiation request message in step S605 does not include a sample alignment requirement, the sample alignment requirement included in the participant discovery request message in step S606 is determined by the VFL server, or the VFL server may negotiate the sample alignment requirement with the primary participant, which is not specifically limited in this embodiment of the present application.

[0313] S607: The VFLSF network element sends a participant discovery response message to the VFL server. Correspondingly, the VFL server receives the participant discovery response message from the VFLSF network element.

[0314] The participant discovery response message includes information about at least one candidate slave participant. The information about the at least one candidate slave participant may include an identifier or address of the at least one candidate slave participant. The identifier or address of the at least one candidate slave participant may be used by the VFL server or the master participant to communicate with the at least one candidate slave participant.

[0315] It should be understood that at least one candidate slave participant may include n candidate slave participants, where n is an integer greater than or equal to 1, for example, n may be 1, 2, 3, or a larger value, and this embodiment of the present application does not specifically limit this.

[0316] In one possible implementation, the information of the at least one candidate slave participant may further include: sample alignment capability of each of the at least one candidate slave participant. The sample alignment capability of each candidate slave participant may include: a sample alignment type supported by each candidate slave participant, a supported sample alignment algorithm, or a time period for which sample alignment is supported.

[0317] It can be understood that for the sample alignment capability, please refer to the relevant description of the aforementioned step S601 for details, which will not be repeated here.

[0318] It should be understood that after the VFL server determines a candidate slave participant that can perform sample alignment with the master participant, the VFL server can trigger the master participant to perform vertical federated learning with at least one candidate slave participant through steps S608 and S609.

[0319] Steps S608 and S609 are described below.

[0320] S608: The VFL server sends a first vertical federated learning request message to at least one candidate slave participant. Correspondingly, the at least one candidate slave participant receives the first vertical federated learning request message from the VFL server.

[0321] The first vertical federated learning request message is used to initiate a vertical federated learning process, or the first vertical federated learning request message is used to indicate the initiation of a vertical federated learning preparation process. In other words, by sending the first vertical federated learning request message to at least one candidate slave participant, the VFL server can notify the at least one candidate slave participant that a vertical federated learning process will be performed next, so that the at least one candidate slave participant can prepare in advance for the relevant configuration of executing the vertical federated learning process.

[0322] It can be understood that the at least one candidate slave participant is a candidate slave participant (or potential slave participant) that can perform sample alignment with the master participant.

[0323] In one possible implementation, the first vertical federated learning request message includes information about the master participant. The information about the master participant may include an identifier or address of the master participant. In other words, the VFL server may notify at least one candidate slave participant capable of performing sample alignment with the master participant of the information about the master participant through the first vertical federated learning request message. It should be understood that at least one candidate slave participant may verify, based on the identifier or address of the master participant, whether a subsequently received message about sample alignment is sent by the master participant, to improve security.

[0324] Exemplarily, the information of the main participant may also include the vertical federated learning group identifier supported by the main participant, the supported vertical federated learning model type identifier, the supported training mode, or the supported sample alignment capability, etc. The supported sample alignment capability may include at least one of the following: supported sample type, supported sample alignment algorithm, or supported sample alignment time period. The embodiments of the present application do not make specific limitations on this.

[0325] Optionally, the first vertical federated learning request message may include a vertical federated learning preparation process identifier (VFL preparation flag), which may be used to indicate the initiation of a vertical federated learning preparation process. It is understood that if the first vertical federated learning request message includes information indicating the initiation of a vertical federated learning preparation process, the first vertical federated learning request message may not carry the vertical federated learning preparation process identifier.

[0326] It can also be understood that the VFL server sending the first vertical federated learning request message to at least one candidate slave participant may mean that the VFL server sends the first vertical federated learning request message to each candidate slave participant in the at least one candidate slave participant.

[0327] S609: The VFL server sends a second vertical federated learning request message to the primary participant. Correspondingly, the primary participant receives the second vertical federated learning request message from the VFL server.

[0328] The second vertical federated learning request message is similar to the first vertical federated learning request message in that the second vertical federated learning request message can be used to indicate the initiation of a vertical federated learning process, or can also be used to indicate the initiation of a vertical federated learning preparation process. In addition, if the second vertical federated learning request message does not include an indication of the initiation of a vertical federated learning preparation process, the second vertical federated learning request message can include a vertical federated learning preparation process identifier.

[0329] It should be understood that the difference between the second vertical federated learning request message and the first vertical federated learning request message is that the second vertical federated learning request message does not include information about the master participant, but instead includes information about at least one candidate slave participant in step S607. The information about the at least one candidate slave participant may include the identifier and / or address of each of the at least one candidate slave participant. Furthermore, the information about the at least one candidate slave participant may also include supported sample alignment capabilities.

[0330] The sample alignment capabilities supported by each of the at least one candidate slave participant may include at least one of the following: supported sample types, supported sample alignment algorithms, or supported sample alignment time periods. The sample alignment capabilities supported by each candidate slave participant can be specifically described in the relevant description of step S601 and are not further described here.

[0331] It will be appreciated that the identifier or address of each candidate slave participant can be used by the master participant to communicate with each candidate slave participant. The sample alignment capabilities supported by each candidate slave participant can be used by the master participant to determine a sample alignment algorithm or time period that is commonly supported by the master participant and at least one candidate slave participant, thereby enabling the master participant to more effectively negotiate a sample alignment algorithm or time period with at least one candidate slave participant.

[0332] It should be understood that after the master participant obtains relevant information of at least one candidate slave participant through the second vertical federated learning request message, the master participant can initiate a sample alignment process. The sample alignment process may include a sample alignment negotiation process and a sample alignment execution process. The sample alignment negotiation process can be used to determine the target slave participant for actually performing sample alignment, the target sample alignment algorithm actually used, and the target time period for actually performing sample alignment. The sample alignment execution process may refer to the process of performing sample alignment between the master participant and the target slave participant, where the target slave participant is a candidate slave participant determined by the master participant from at least one candidate slave participant.

[0333] It can be understood that the sample alignment negotiation process is only an exemplary name, and can also be called a sample alignment auxiliary process, or other names, and the embodiments of the present application do not specifically limit this.

[0334] The following introduces steps S610 to S612, which are the sample alignment negotiation process.

[0335] S610: The master participant sends a sample alignment request message to at least one candidate slave participant. Correspondingly, the at least one candidate slave participant receives the sample alignment request message from the master participant.

[0336] The sample alignment request message may be used to request sample alignment, or the sample alignment request message may be used to negotiate a sample alignment method, which may include a sample alignment type, a sample alignment algorithm, or a time period for sample alignment.

[0337] In one possible implementation, the sample alignment request message includes first information indicating a first common sample alignment algorithm supported by the master and at least one candidate slave. That is, the master can determine the first common sample alignment algorithm supported by the master and at least one candidate slave, and send the sample alignment request message to the at least one candidate slave to indicate that sample alignment can be performed using the first common sample alignment algorithm. This can reduce the probability that at least one candidate slave does not support sample alignment, thereby improving the success rate and efficiency of sample alignment negotiation.

[0338] For example, the master participant can determine the sample alignment algorithm supported by each candidate slave participant in at least one candidate slave participant based on the second vertical federated learning request message in step S609, and take the intersection between the sample alignment algorithm supported by the master participant and the sample alignment algorithm supported by each candidate slave participant, thereby obtaining the intersection including the first common sample alignment algorithm.

[0339] In addition, the aforementioned intersection may include one or more first common sample alignment algorithms. Exemplarily, for the convenience of expression, the symbol "01" represents PSI based on Diffie-Hellman, the symbol "02" represents PSI based on homomorphic encryption, the symbol "03" represents PSI based on RSA and hash algorithm, and the symbol "04" represents PSI based on OT-PRF. Assume that at least one candidate slave participant includes candidate slave participant 1 and candidate slave participant 2, the sample alignment algorithms supported by candidate slave participant 1 are [01, 02, 03], the sample alignment algorithms supported by candidate slave participant 2 are [02, 03, 04], and the sample alignment algorithms supported by the master participant are [01, 02, 03, 04]. In this way, the intersection between the master participant, candidate slave participant 1, and candidate slave participant 2 is [02, 03], that is, the intersection includes 2 first common sample alignment algorithms.

[0340] It should be understood that if the aforementioned intersection contains only one sample alignment algorithm, the first information may indicate the one sample alignment algorithm; if the aforementioned intersection contains multiple sample alignment algorithms, the first information may indicate the multiple sample alignment algorithms, or indicate at least one of the multiple sample alignment algorithms (i.e., a subset of the intersection), and the embodiments of the present application do not make specific limitations on this.

[0341] It can be understood that the main participant can select at least one of the sample alignment algorithms from multiple sample alignment algorithms that has less computing power requirements for the candidate slave participants (for example, the computing power requirements for participant A in Example 2 are smaller). This can reduce the probability that the candidate slave participants do not support sample alignment due to large computing power requirements, thereby improving the success rate and efficiency of sample alignment negotiation.

[0342] In addition, the name of the first common sample alignment algorithm is only an example. The first common sample alignment algorithm may also be called an available sample alignment algorithm (available sample alignment algorithms) or other names, which is not specifically limited in the embodiments of the present application.

[0343] In a possible implementation, the first information is further used to indicate at least one of the following: the number of samples supported by the main participant for sample alignment, the sample alignment type supported by the main participant, or at least one time period for which the main participant supports sample alignment.

[0344] Among them, the number of samples supported by the master participant for sample alignment can be used by the candidate slave participant to determine the computing power requirements of the sample alignment algorithm. For example, for a sample alignment algorithm with the same computing power requirements for each participant (such as the sample alignment algorithm based on Diffie-Hellman's PSI in Example 1), the computing power requirements of the sample alignment algorithm are not only related to the number of samples of the candidate slave participant itself, but also to the number of samples of the master participant. Then, the candidate slave participant can determine the computing power requirements of the first common sample alignment algorithm based on its own number of samples and the number of samples of the master participant, so that the candidate slave participant can determine whether its computing power can support the use of the first common sample alignment algorithm to perform sample alignment, thereby improving the success rate of sample alignment.

[0345] In addition, the number of samples supported by the primary participant for sample alignment can refer to: the number of all samples in the primary participant's dataset used for longitudinal federated learning model training, or a subset of the number of all samples in the dataset. It is understood that a subset of the number of all samples can be used to indicate that the primary participant expects to use the number of samples corresponding to the subset for sample alignment.

[0346] It is understood that, if the first common sample alignment algorithm is a sample alignment algorithm that requires the same computing power for each participant, the primary participant indicates the number of samples supported by the primary participant for sample alignment through the first information. Alternatively, if the first common sample alignment algorithm is a sample alignment algorithm that requires different computing power for each participant, the primary participant indicates the number of samples supported by the primary participant for sample alignment through the first information.

[0347] The sample alignment type supported by the main participant may be a sample alignment type that the main participant can support, or may be used to indicate the sample alignment type that the main participant expects to perform. For example, the sample alignment type supported by the main participant may be one that supports two-party sample alignment and / or multi-party sample alignment. It can be understood that for a sample alignment algorithm that can simultaneously support two-party sample alignment and multi-party sample alignment, the computing power requirement of the sample alignment algorithm for two-party sample alignment may be different from the computing power requirement for multi-party sample alignment. Furthermore, the sample alignment type supported by the main participant can assist the candidate slave participant in determining the computing power requirement of the sample alignment algorithm, so that the candidate slave participant can determine whether to support sample alignment.

[0348] At least one time period in which the master participant supports sample alignment can be used by the candidate slave participant to determine whether to support sample alignment. For example, the candidate slave participant can determine whether to support sample alignment based on the at least one time period and the load of the candidate slave participant in the at least one time period. For example, if the load of the candidate slave participant in the at least one time period is less than a threshold value 1, the candidate slave participant can provide sufficient local resources for sample alignment in the at least one time period, that is, the candidate slave participant determines that it supports sample alignment in the at least one time period. For another example, if the load of the candidate slave participant in the at least one time period is greater than a threshold value 2, the local resources that the candidate slave participant can call in the at least one time period do not match the computing power requirements of the sample alignment algorithm, that is, the candidate slave participant determines that it cannot support sample alignment in the at least one time period.

[0349] It can be understood that the specific values ​​of the above threshold 1 and threshold 2 depend on the actual implementation, and the embodiments of the present application do not make specific limitations on this.

[0350] Exemplarily, local resources may include computing resources, storage resources, and network resources, wherein computing resources may include processor frequency, storage resources may include memory size and storage capacity of local memory, and network resources may include throughput.

[0351] Optionally, at least one first common sample alignment algorithm may be associated with at least one time period.

[0352] It can be understood that, considering that the load of the master participant is different in different time periods, the computing power supported by the local resources that can be called in different time periods is also different, that is, the first common sample alignment algorithm supported by the master participant in different time periods may be different. In other words, since the first common sample alignment algorithm supported by the master participant in different time periods may be different, by associating at least one first common sample alignment algorithm with at least one time period, the candidate slave participant can determine the first common sample alignment algorithm supported by the master participant in different time periods, and determine whether the computing power requirements of the first common sample alignment algorithm corresponding to the different time periods are supported based on the local resources that can be called in different time periods, so as to further determine whether to support the execution of sample alignment.

[0353] For example, the first information indicates the first common sample alignment algorithm 1 and the first common sample alignment algorithm 2, at least one time period includes time period 1 and time period 2, the first common sample alignment algorithm 1 is associated with time period 1, which may indicate that the main participant can support the use of the first common sample alignment algorithm 1 to perform sample alignment in time period 1, and the first common sample alignment algorithm 2 is associated with time period 2, which may indicate that the main participant can support the use of the first common sample alignment algorithm 2 to perform sample alignment in time period 2.

[0354] In addition, a first common sample alignment algorithm can be associated with at least two time periods, for example, the first common sample alignment algorithm 1 is associated with time period 1 and time period 2 respectively, to indicate that the main participant supports the use of the first common sample alignment algorithm 1 to perform sample alignment in time period 1 and time period 2.

[0355] Alternatively, a time period can be associated with at least two first common sample alignment algorithms, such as time period 1 is associated with first common sample alignment algorithm 1 and first common sample alignment algorithm 2, to indicate that the main participant supports the use of first common sample alignment algorithm 1 or first common sample alignment algorithm 2 to perform sample alignment in time period 1.

[0356] It should be understood that the number of samples supported by the above-mentioned main participant for sample alignment, the type of sample alignment supported by the main participant, or at least one time period for which the main participant supports sample alignment can be indicated using parameters other than the first information, and the embodiments of the present application do not specifically limit this.

[0357] S611: At least one candidate slave participant sends a sample alignment response message to the master participant. Correspondingly, the master participant receives the sample alignment response message from the at least one candidate slave participant.

[0358] The sample alignment response message may be used to indicate support for sample alignment, and / or a second common sample alignment algorithm, where the second common sample alignment algorithm is a sample alignment algorithm determined from a plurality of first common sample alignment algorithms; or, the sample alignment response message may be used to indicate non-support for sample alignment, and / or the reason why sample alignment is not supported.

[0359] It will be appreciated that at least one candidate slave participant sending a sample alignment response message to the master participant may include: each of the at least one candidate slave participant sending a sample alignment response message to the master participant. Accordingly, the master participant receiving a sample alignment response message from at least one candidate slave participant may include: the master participant receiving a sample alignment response message from each of the at least one candidate slave participant.

[0360] It should be understood that each of the at least one candidate slave participant can determine whether to support sample alignment using the first common sample alignment algorithm indicated by the first information based on the sample alignment request message and the local information of each candidate slave participant. The local information of the candidate slave participant may include the usage of local resources or available local resources. For details about local resources, please refer to the description of local resources in step S610 and will not be repeated here.

[0361] In addition, if the candidate slave participant supports using the first common sample alignment algorithm to perform sample alignment, the second common sample alignment algorithm indicated by the candidate slave participant may be the first common sample alignment algorithm.

[0362] It can be understood that the candidate slave participant determines whether to support sample alignment based on the sample alignment request message and local information. The specific principle can be found in the relevant description of the first information in step S610. Two examples are listed below for further explanation.

[0363] Exemplarily, for the first common sample alignment algorithm being the PSI based on RSA and hash algorithms in Example 2 above, the slave participant can determine the computing power requirements of the first common sample alignment algorithm based on the first common sample alignment algorithm and the number of samples contained in the data set of the slave participant for longitudinal federated learning model training, and determine whether to support the use of the first common sample alignment algorithm to perform sample alignment based on whether the available local resources meet the computing power requirements. In addition, when the first information also indicates at least one time period in which the master participant supports sample alignment, the slave participant can also estimate the load corresponding to the at least one time period, determine whether the available local resources corresponding to the at least one time period meet the computing power requirements, and then determine whether to support the use of the first common sample alignment algorithm to perform sample alignment within the at least one time period. If the slave participant determines that it supports it, then the first common sample alignment algorithm is the second common sample alignment algorithm; if not, then the first common sample alignment algorithm is not the second common sample alignment algorithm.

[0364] Exemplarily, for the first common sample alignment algorithm being the Diffie-Hellman-based PSI in Example 1 above, the slave party can determine the computing power requirements of the first common sample alignment algorithm based on the number of samples supported by the master party for sample alignment indicated by the first information, and the number of samples contained in the slave party's dataset for training the longitudinal federated learning model, and determine whether to support the use of the first common sample alignment algorithm to perform sample alignment based on whether the available local resources meet the computing power requirements. If the slave party determines that it supports the use of the first common sample alignment algorithm to perform sample alignment, then the first common sample alignment algorithm is the second common sample alignment algorithm; if not, then the first common sample alignment algorithm is not the second common sample alignment algorithm. In addition, when the first information also indicates at least one time period in which the master participant supports sample alignment, the slave participant can also estimate the load corresponding to the at least one time period, determine whether the callable local resources corresponding to the at least one time period meet the above-mentioned computing power requirements, and then determine whether to support the use of the first common sample alignment algorithm to perform sample alignment within the at least one time period. If the slave participant determines to support it, the first common sample alignment algorithm is the second common sample alignment algorithm; if not, the first common sample alignment algorithm is not the second common sample alignment algorithm.

[0365] It can be understood that when the first information indicates multiple first common sample alignment algorithms, the participating party can traverse each first common sample alignment algorithm in the multiple first common sample alignment algorithms according to the above example of determining whether to use the first common sample alignment algorithm to perform sample alignment, and determine whether to support the execution of sample alignment.

[0366] In addition, there may be multiple second common sample alignment algorithms. For example, the multiple first common sample alignment algorithms are: sample alignment algorithm 1, sample alignment algorithm 2, and sample alignment algorithm 3, and the second common sample alignment algorithms may be sample alignment algorithm 2 and sample alignment algorithm 3.

[0367] It should be understood that the above implementation of the determination by the participating party whether to support the execution of sample alignment is only an example. The participating party can use other methods to make judgments based on the first information. For specific implementation, please refer to step S610. The embodiment of this application does not make specific limitations on this.

[0368] The following describes the sample alignment response message in two cases based on whether the participating parties support the execution of sample alignment.

[0369] Case 1: Sample alignment is performed with support from participating parties.

[0370] For a slave party supporting sample alignment, the sample alignment response message may be used to indicate support for sample alignment and / or a second common sample alignment algorithm.

[0371] It can be understood that for the first information in step S610, only one first common sample alignment algorithm is indicated. In order to save network overhead, the sample alignment response message may only indicate support for sample alignment. For the first information indicating multiple first common sample alignment algorithms, if the slave party supports the multiple first sample alignment sample algorithms, in order to save network overhead, the sample alignment response message may only indicate support for sample alignment; if the slave party supports the second common sample alignment algorithm among the multiple first common sample alignment algorithms, in order to save network overhead, the sample alignment response message may only indicate the second common sample alignment algorithm, to indicate that the slave party supports the use of the second common sample alignment algorithm among the multiple first common sample alignment algorithms to perform sample alignment. Of course, the sample alignment response message may also indicate support for the execution of sample alignment and the second common sample alignment algorithm, which is not specifically limited in the embodiments of the present application.

[0372] Optionally, the sample alignment response message may also be used to indicate: a time period for which the slave party supports sample alignment, and / or a number of samples supported by the slave party for sample alignment.

[0373] It is understood that the time period during which the slave participant supports sample alignment can be used by the master participant to determine the time period during which the slave participant can perform sample alignment, so that the master participant can determine whether to select the slave participant to perform sample alignment based on the time period. The time period during which the slave participant supports sample alignment can include one or more time periods, and the one or more time periods can be associated with the second common sample alignment algorithm. For example, the time period during which the slave participant supports sample alignment includes time period 3 and time period 4, and the multiple second common sample alignment algorithms include sample alignment algorithm 1 and sample alignment algorithm 2. Sample alignment algorithm 1 can be associated with time period 3 to indicate that the slave participant supports the use of sample alignment algorithm 1 to perform sample alignment in time period 3. Similarly, sample alignment algorithm 2 can be associated with time period 4 to indicate that the slave participant supports the use of sample alignment algorithm 2 to perform sample alignment in time period 4. Of course, sample alignment algorithm 1 can be associated with time period 3 and time period 4 respectively to indicate that the slave participant supports the use of sample alignment algorithm 1 to perform sample alignment in time period 3 and time period 4.

[0374] In addition, when the first information in step S610 also indicates at least one time period for which the main participant supports sample alignment (the at least one time period may be associated with the first common sample alignment algorithm), the time period for which the above-mentioned slave participant supports sample alignment may be one or more time periods in the at least one time period, or a combination of a subset of one or more time periods in the at least one time period.

[0375] For example, assuming that the primary participant supports at least one time period for sample alignment including: [9:00-12:00] and [14:00-18:00], then the time period for the secondary participant to support sample alignment may be one or more time periods in the at least one time period, for example, [9:00-12:00], or [14:00-18:00], or [9:00-12:00] and [14:00-18:00]; or, the secondary participant may support at least one time period for sample alignment. The time period aligned with the square support sample can be a subset of one or more time periods in the at least one time period, for example, it can be a subset of [9:00-12:00] [9:30-11:00], or a subset of [14:00-18:00] [15:00-17:30], or a subset of [9:45-11:25] of [9:00-12:00] and a subset of [14:30-17:45] of [14:00-18:00], etc.

[0376] The number of samples supported by the slave participants for sample alignment can be used by the master participant to determine the computing power requirements of the second common sample alignment algorithm. It can be understood that for the second common sample alignment algorithm, which is a sample alignment algorithm with the same computing power requirements for each participant (such as the Diffie-Hellman-based PSI in Example 1), the computing power requirements of the sample alignment algorithm are not only related to the number of samples of the master participant itself, but also to the number of samples of the slave participants. Then, the master participant can determine the computing power requirements of the second common sample alignment algorithm based on its own number of samples and the number of samples of the slave participants, so as to determine whether the sample alignment can be supported in the actual execution of the sample alignment, thereby improving the success rate of the sample alignment.

[0377] In addition, the number of samples supported by the slave party for sample alignment may refer to: the number of samples of all samples in the slave party's data set for longitudinal federated learning model training, or a subset of the number of samples of all samples in the data set. It can be understood that a subset of the number of samples of all samples can be used to indicate that the slave party expects to use the samples corresponding to the subset for longitudinal federated learning model training. For example, if the slave party determines that the local computing power cannot support the use of the second common sample alignment algorithm to align all samples, the slave party may use the samples corresponding to the subset of all samples for sample alignment and subsequent longitudinal federated learning model training.

[0378] For example, assuming that the first information indicates two first common sample alignment algorithms, namely sample alignment algorithm 1 and sample alignment algorithm 2, the first information also indicates that the number of samples supported by the main participant for sample alignment is 1 billion, and the slave participant determines that the number of samples contained in the data set for training the longitudinal federated learning model is 100,000. Based on the above conditions, the slave participant determines that it can support sample alignment algorithm 1 but cannot support sample alignment algorithm 2. In this way, the sample alignment response message may include: sample alignment algorithm 1 (i.e., the second common sample alignment algorithm), the number of samples supported by the slave participant for sample alignment is 100,000, and the time period for the slave participant to support the use of sample alignment algorithm 1 to perform sample alignment.

[0379] Case 2: The slave party does not support performing sample alignment.

[0380] In the case where the slave party does not support sample alignment, the sample alignment response message may be used to indicate that sample alignment is not supported and / or the reason why sample alignment is not supported.

[0381] It can be understood that after the slave participant determines that it does not support the execution of sample alignment, it can send a response message to the master participant to indicate that it does not support the execution of sample alignment, so as to inform the master participant that the slave participant cannot support the execution of sample alignment. Alternatively, after the slave participant determines that it does not support the execution of sample alignment, the slave participant can send the reason for not supporting the execution of sample alignment to assist the master participant in adjusting the first common sample alignment algorithm indicated by the first information, or the number of samples supported by the master participant for sample alignment, or the time period for which the master participant supports sample alignment, when the master participant subsequently initiates the sample alignment request again, so as to improve the success rate of the relevant configurations of the negotiated sample alignment (such as the sample alignment algorithm, the number of samples, or the time period for sample alignment, etc.).

[0382] In addition, the reasons for not supporting sample alignment may include one or more reasons. For example, computing power is not supported, or the time period is not supported. Among them, computing power not supported may include the number of samples supported by the master participant for sample alignment being too large, or the number of samples corresponding to the data set used by the slave participant for training the longitudinal federated learning model being too large. It is understood that the above reasons for not supporting sample alignment are merely examples and may include other reasons, which are not specifically limited in the embodiments of the present application.

[0383] Exemplarily, the sample alignment response message may include a cause value parameter (or a failure indication with cause code), where the cause value is used to indicate why sample alignment is not supported. It will be appreciated that indicating the reason for not performing sample alignment using the cause value can reduce indication overhead and improve reliability.

[0384] For example, the protocol can pre-configure reason values. For example, the pre-configured symbol "01" is used to indicate that the reason for not supporting sample alignment is that the computing power is not supported, and "02" is used to indicate that the reason for not supporting sample alignment is that the time period is not supported. For another example, the pre-configured symbol "01" is used to indicate that the reason for not supporting sample alignment is that the number of samples supported by the primary participant for sample alignment is too large, and "02" is used to indicate that the reason for not supporting sample alignment is that the time period is not supported.

[0385] It can be understood that the above protocol pre-configured cause values ​​are only examples. The symbols corresponding to the cause values ​​can also be negotiated in advance between the slave participants, the main participants, and the VFLSF network elements. The symbols corresponding to the cause values ​​can also be represented in other ways (such as bits). The embodiments of the present application do not make specific limitations on this.

[0386] In addition, the sample alignment response message may include two parameters, namely a failure indication and a reason value. The failure indication is used to indicate that the slave party does not support the execution of sample alignment, and the reason value indicates the reason why the execution of sample alignment is not supported. The embodiment of the present application does not make specific limitations on this.

[0387] Optionally, the sample alignment response message may be used to indicate the maximum number of samples that the primary participant is expected to use.

[0388] It is understood that the maximum number of samples expected to be used by the master participant can be used to indicate the maximum number of samples that the slave participant can accept when performing sample alignment based on local resources. If the master participant can perform sample alignment using a dataset that is less than or equal to the maximum number of samples expected to be used by the master participant, the slave participant can support performing sample alignment.

[0389] In addition, when the first information indicates multiple first common sample alignment algorithms, the sample alignment response message can also be used to indicate the first common sample alignment algorithm associated with the maximum number of samples expected to be used by the main participant. In this way, the main participant can determine which sample alignment algorithm corresponds to the maximum number of samples expected to be used by the main participant indicated in the sample alignment response message, and then assist the main participant in adjusting the number of samples supported by the main participant for sample alignment corresponding to the first common sample alignment algorithm when initiating a sample alignment request again (i.e., step S610) subsequently.

[0390] For example, assuming that the first information indicates two first common sample alignment algorithms, namely sample alignment algorithm 1 and sample alignment algorithm 2, the first information also indicates that the number of samples supported by the main participant for sample alignment is 1 billion, and the number of samples contained in the data set determined by the slave participant for training the longitudinal federated learning model is 100,000. Based on the above conditions, the slave participant determines that the computing power does not support the number of samples supported by the main participant for sample alignment because the number of samples supported by the main participant for sample alignment is too large. If the number of samples supported by the main participant for sample alignment is 500 million, the slave participant can support sample alignment algorithm 1; if the number of samples supported by the main participant for sample alignment is 700 million, the slave participant can support sample alignment algorithm 2, and the sample alignment response message may include: a failure indication, a reason value (for example, the number of samples supported by the main participant for sample alignment is too large), the first maximum number of samples expected to be used by the main participant associated with sample algorithm 1 is 500 million, and the second maximum number of samples expected to be used by the main participant associated with sample algorithm 2 is 700 million.

[0391] In addition, the above is only an example. The sample alignment response message can be a combination of one or more of a failure indication, a reason value, and the maximum number of samples that the slave participant expects the master participant to use. The embodiments of the present application do not specifically limit this.

[0392] It should be understood that in the case where the slave party does not support sample alignment, the sample alignment response message may also include other information such as the time period in which the slave party expects to perform sample alignment, and the embodiments of the present application do not specifically limit this.

[0393] It can be understood that if the sample alignment response information of each candidate slave participant in at least one candidate slave participant indicates that sample alignment is not supported, the sample alignment negotiation fails, and the master participant can re-initiate the sample alignment negotiation process. For example, the master participant can adjust the first common sample alignment algorithm, the number of samples supported by the master participant for sample alignment, and one or more of the at least one time period for which the master participant supports sample alignment, and send the adjusted sample alignment request message to improve the success rate of the relevant configuration of the negotiated sample alignment. In addition, if the sample alignment negotiation fails again, the master participant can also re-initiate the sample alignment negotiation process until the sample alignment negotiation succeeds, or the number of repeated executions of the sample alignment negotiation process reaches an upper limit. The upper limit may be pre-configured by the protocol, or negotiated in advance by at least two of the master participant, the slave participant, the VFL server, or the VFLSF network element. This embodiment of the present application does not specifically limit this.

[0394] It can also be understood that if the sample alignment response messages of at least some of the at least one candidate slave participants indicate support for sample alignment, the master participant can execute step S612, which is used to indicate the target sample alignment algorithm to at least one target slave participant that meets the sample alignment execution conditions among the at least some of the slave participants. In addition, the master participant can also indicate not to perform sample alignment to the slave participants that do not meet the sample alignment execution conditions among the at least some of the slave participants, so as to inform the slave participants that do not meet the sample alignment execution conditions that sample alignment will not be performed next.

[0395] Step S612 is described below.

[0396] S612: The master participant sends a first sample alignment negotiation result message to the target slave participant. Accordingly, the target slave participant receives the first sample alignment negotiation result message from the master participant. The first sample alignment negotiation result message indicates the target sample alignment algorithm.

[0397] It can be understood that the main participant can determine the target slave participant and the target sample alignment algorithm based on the sample alignment response message of at least one candidate slave participant. For example, in the case where the sample alignment response message of at least one candidate slave participant indicates that there is only one slave participant that supports the execution of sample alignment, the target slave participant is the slave participant that supports the execution of sample alignment, and the target sample alignment algorithm is the second common sample alignment algorithm indicated in the sample alignment response message that indicates support for the execution of sample alignment. In addition, in the case where the sample alignment response message of at least one candidate slave participant indicates that there is only one slave participant that supports the execution of sample alignment, and the sample alignment response message of the slave participant indicates multiple second common sample alignment algorithms, the main participant can determine a second common sample alignment algorithm from the multiple second common sample alignment algorithms as the target sample alignment algorithm. Among them, there can be multiple ways for the main participant to determine an implementation of a second common sample alignment algorithm from multiple second common sample alignment algorithms. For example, the main participant randomly selects a second common sample alignment algorithm from multiple second common sample alignment algorithms, or the main participant selects the second common sample alignment algorithm with the smallest computing power requirement (i.e., computational complexity) from multiple second common sample alignment algorithms, etc. The embodiments of the present application do not make specific limitations on this.

[0398] In addition, if the sample alignment response message of at least one candidate slave participant indicates that there are multiple slave participants supporting sample alignment, the master participant may determine the target slave participant and the target sample alignment algorithm from the multiple slave participants based on the sample alignment response messages of the multiple slave participants. For example, if the sample alignment response message of at least one candidate slave participant is the sample alignment response message of at least one candidate slave participant in step S611, the master participant may determine the multiple slave participants supporting sample alignment based on the sample alignment response message of the at least one candidate slave participant, and determine the target slave participant and the target sample alignment algorithm based on the intersection of the second common sample alignment algorithms indicated by the multiple participants.

[0399] That is to say, the master participant can determine the slave participants that support the sample alignment based on the sample alignment response message of at least one candidate slave participant, and further determine the target sample alignment algorithm used to perform the sample alignment based on multiple slave participants that support the sample alignment, as well as the target slave participants that meet the computing power requirements of the target sample alignment algorithm, so as to avoid the failure of sample alignment due to failure to meet the computing power requirements of the sample alignment algorithm during the sample alignment process, thereby improving the success rate and efficiency of the sample alignment.

[0400] It should be understood that the master participant may determine the target slave participant and the target sample alignment algorithm according to the sample alignment response message of at least one candidate slave participant in various ways, which are described below.

[0401] Example A: The master participant determines a combination of multiple slave participants supporting the same sample alignment algorithm based on a sample alignment response message from at least one candidate slave participant, and selects the combination with the largest number of supported slave participants as the target sample alignment algorithm and target slave participant.

[0402] For example, at least one candidate slave participant in step S611 includes: candidate slave participants 1 to 5, the second common sample alignment algorithm indicated by the sample alignment response message of candidate slave participant 1 is: sample alignment algorithm 1 and sample alignment algorithm 2, the second common sample alignment algorithm indicated by the sample alignment response message of candidate slave participant 2 is: sample alignment algorithm 2 and sample alignment algorithm 3, the sample alignment response message of candidate slave participant 3 indicates that sample alignment is not supported, the second common sample alignment algorithm indicated by the sample alignment response message of candidate slave participant 4 is: sample alignment algorithm 3 and sample alignment algorithm 4, and the second common sample alignment algorithm indicated by the sample alignment response message of candidate slave participant 5 is: sample alignment algorithm 2, and then in the intersection of the second common sample alignment algorithms of the above-mentioned different participants, the combination of multiple participants supporting the same sample alignment algorithm includes: candidate slave participant 1, candidate slave participant 2, and candidate slave participant 5 support sample alignment algorithm 2, and candidate slave participant 2 and candidate slave participant 4 support sample alignment algorithm 3.

[0403] Based on the above two combinations, the master participant can select the combination with the largest number of supported candidate slave participants as the target sample alignment algorithm and target slave participants. That is, the master participant can determine that the target slave participants are: candidate slave participant 1, candidate slave participant 2, and candidate slave participant 5, and the target sample alignment algorithm is sample alignment algorithm 2. Furthermore, the master participant can send a first sample alignment negotiation result message to candidate slave participant 1, candidate slave participant 2, and candidate slave participant 5, respectively. The first sample alignment negotiation result message is used to indicate that the target sample alignment algorithm for performing sample alignment is sample alignment algorithm 2.

[0404] It can be understood that the purpose of vertical federated learning model training is to improve the accuracy of model training by expanding the number of features. In the sample alignment process, the main participant can increase the number of expanded features and thus improve the accuracy of model training by determining more targets for sample alignment from the participants.

[0405] Example B: When the combination with the largest number of supported candidate slave participants is at least two different combinations, the master participant may randomly select one of the combinations to determine the target slave participant and the target sample alignment algorithm.

[0406] For example, assume that the difference between the sample alignment response message of at least one candidate slave participant between Example B and Example A is that the second common sample alignment algorithm indicated by the sample alignment response message of candidate slave participant 5 in Example B is: sample alignment algorithm 2 and sample alignment algorithm 3, that is, in the intersection of the second common sample alignment algorithms of the above-mentioned different participants, the combination of multiple participants supporting the same sample alignment algorithm includes: candidate slave participant 1, candidate slave participant 2, and candidate slave participant 5 support sample alignment algorithm 2, candidate slave participant 2, candidate slave participant 4, and candidate slave participant 5 support sample alignment algorithm 3.

[0407] Since the two combinations above correspond to the same number of candidate slave participants, the master participant can randomly select one of the combinations to determine the target slave participant and the target sample alignment algorithm. For example, the master participant can send a first sample alignment negotiation result message to candidate slave participant 1, candidate slave participant 2, and candidate slave participant 5, respectively, where the first sample alignment negotiation result message indicates that the target sample alignment algorithm is sample alignment algorithm 2; or the master participant can send a first sample alignment negotiation result message to candidate slave participant 2, candidate slave participant 4, and candidate slave participant 5, respectively, where the first sample alignment negotiation result message indicates that the target sample alignment algorithm is sample alignment algorithm 3.

[0408] Example C: The master participant may determine a target slave participant and a target sample alignment algorithm based on a range of the number of slave participants performing sample alignment and a sample alignment response message from at least one candidate slave participant.

[0409] It is understood that the range of the number of slave participants performing sample alignment may include: the maximum number of slave participants performing sample alignment, and / or the minimum number. For example, the maximum number of slave participants performing sample alignment can be indicated by a first threshold, and then the range of the number of slave participants performing sample alignment is [1, first threshold]. For another example, the minimum number of slave participants performing sample alignment can be indicated by a second threshold, and then the range of the number of slave participants performing sample alignment is [second threshold, +∞). For another example, the minimum number of slave participants performing sample alignment is the second threshold, and the maximum number is the first threshold, and then the range of the number of slave participants performing sample alignment is [second threshold, first threshold].

[0410] For example, assume that the difference between the sample alignment response messages of at least one candidate slave participant between Example C and Example B is that the second common sample alignment algorithm indicated by the sample alignment response message of candidate slave participant 1 in Example C is: sample alignment algorithms 1 to 3, and then in the intersection of the second common sample alignment algorithms of the above-mentioned different participants, multiple combinations supporting the same sample alignment algorithm include: candidate slave participants 1, 2, 4, and 5 support sample alignment algorithm 3, and candidate slave participants 1, 2, and 5 support sample alignment algorithms 2 to 3.

[0411] Taking the example that the range of the number of slave participants performing sample alignment includes a minimum number of 4, the master participant can select candidate slave participants 1, 2, 4, and 5 to support sample alignment algorithm 3 as the target sample alignment algorithm, that is, the master participant can send a first sample alignment negotiation result message to candidate slave participants 1, 2, 4, and 5 respectively, and the first sample alignment negotiation result message indicates that the target sample alignment algorithm is sample alignment algorithm 3.

[0412] Taking the case where the range of the number of slave participants performing sample alignment includes a minimum number of 3 as an example, the master participant may determine the target slave participant and the target sample alignment algorithm according to any one of the above two combinations.

[0413] Taking the example of a maximum number of 3 slave participants performing sample alignment, the master participant can determine the target slave participant and target sample alignment algorithm based on the support of sample alignment algorithms 2 to 3 by candidate slave participants 1, 2, and 5 in the above combination. For example, the target slave participants are candidate slave participants 1, 2, and 5, and the sample alignment algorithm can be sample alignment algorithm 2 and / or sample alignment algorithm 3.

[0414] Taking the example that the number of slave participants performing sample alignment ranges from a minimum of 4 to a maximum of 6, the master participant can select candidate slave participants 1, 2, 4, and 5 to support sample alignment algorithm 3 as the target sample alignment algorithm.

[0415] It can be understood that the main participant can determine the target slave participants and target sample alignment algorithm that meet the sample number requirements for performing sample alignment based on the number range of slave participants performing sample alignment, so as to further avoid the problem of excessive computing power requirements due to too many target slave participants participating in sample alignment, which affects the efficiency and success rate of sample alignment, thereby improving the success rate and efficiency of sample alignment.

[0416] It should be understood that the above examples A to C are examples based on the fact that the sample alignment response message indicates the second common sample alignment algorithm. When the sample alignment response message of the candidate slave participant only indicates support for performing sample alignment, the master participant may assume that the candidate slave participant supports the first common sample alignment algorithm indicated by the sample alignment request message in step S610.

[0417] In addition, on the basis that the sample alignment response message indicates the second common sample alignment algorithm, it also indicates the time period for which the slave participant supports sample alignment, and / or the number of samples supported by the slave participant for sample alignment, the master participant can determine the target slave participant and the target sample alignment algorithm based on the second common sample alignment algorithm indicated by each candidate slave participant, the time period for which the slave participant supports sample alignment, or the number of samples supported by the slave participant for sample alignment, as explained below.

[0418] Example D: The master participant may determine a target slave participant and a target sample alignment algorithm based on the second common sample alignment algorithm indicated by a sample alignment response message of at least one candidate slave participant and the time period during which the slave participant supports sample alignment.

[0419] For example, based on the determination in Example A that the target sample alignment algorithm is sample alignment algorithm 2, it is assumed that the difference between the sample alignment response messages of at least one candidate slave participant between Example D and Example A is that: the sample alignment response message of candidate slave participant 1 also indicates time period 1 associated with sample alignment algorithm 2: [14:00-17:00], the sample alignment response message of candidate slave participant 2 also indicates time period 2 associated with sample alignment algorithm 2: [9:00-11:00], and the sample alignment response message of candidate slave participant 5 also indicates time period 3 associated with sample alignment algorithm 2: [15:00-17:00]. Then, the master participant further determines, based on whether there is an intersection among the time periods of candidate slave participants 1, 2, and 5, that time period 2 does not overlap with time period 1 and time period 3, respectively, that is, candidate slave participant 2 is not the target slave participant, that is, the master participant determines that the target slave participants are candidate slave participants 1 and 5, the target sample alignment algorithm is sample alignment algorithm 2, and the target time period for sample alignment is time period 3.

[0420] In addition, assuming time period 2: [16:30-17:00], there is an intersection between the time periods of candidate slave participants 1, 2, and 5 used to support sample alignment, but the time period of the intersection is only [16:30-17:00], that is, the time period of the intersection is too short, and the master participant can also determine that candidate slave participant 2 is not suitable as the target slave participant.

[0421] It should be understood that, assuming that the time period indicated by the sample response message of each candidate slave participant is not a subset of at least one time period supported by the master participant, the master participant should first determine the intersection between the time period indicated by each candidate slave participant and at least one time period, and determine the target slave participant and target time period based on the intersection.

[0422] Example E: The master participant may determine a target slave participant and a target sample alignment algorithm based on the second common sample alignment algorithm indicated by the sample alignment response message of at least one candidate slave participant and the number of samples supported by the slave participant for sample alignment.

[0423] For example, assume that the difference between the sample alignment response messages of at least one candidate slave participant in Example E and Example A is that the sample alignment response message of candidate slave participant 1 further indicates that the number of samples supported by candidate slave participant 1 for sample alignment is 100,000, the sample alignment response message of candidate slave participant 2 further indicates that the number of samples supported by candidate slave participant 2 for sample alignment is 500,000, and the sample alignment response message of candidate slave participant 5 further indicates that the number of samples supported by candidate slave participant 5 for sample alignment is 2 million. If the master participant determines that the computing power requirement of sample alignment algorithm 2 is not met when the number of samples of candidate slave participant 5 is 2 million, the master participant determines that candidate slave participant 5 is not the target slave participant, that is, the target slave participants are candidate slave participants 1 and 2 respectively, and the target sample alignment algorithm is sample alignment algorithm 2.

[0424] Example F: The master participant can determine the target slave participant and the target sample alignment algorithm based on the second common sample alignment algorithm indicated by the sample alignment response message of at least one candidate slave participant, the time period for the slave participant to support sample alignment, and the number of samples supported by the slave participant for sample alignment.

[0425] For example, based on the determination in Example A that the target sample alignment algorithm is sample alignment algorithm 2, it is assumed that the difference between Example F and Example A in the sample alignment response message of at least one candidate slave participant is that: the sample alignment response message of candidate slave participant 1 also indicates the time period 1 associated with sample alignment algorithm 2: [14:00-15:00], and the number of samples supported by candidate slave participant 1 for sample alignment is 2 million; the sample alignment response message of candidate slave participant 2 also indicates the time period 2 associated with sample alignment algorithm 2: [14:00-17:00], and the number of samples supported by candidate slave participant 2 for sample alignment is 100,000; the sample alignment response message of candidate slave participant 5 also indicates the time period 3 associated with sample alignment algorithm 2: [15:00-17:00], and the number of samples supported by candidate slave participant 2 for sample alignment is 100,000. Party 5 supports 500,000 samples for sample alignment; then the master participant determines that candidate slave participants 1, 2, and 5 jointly support sample alignment algorithm 2, the intersection of time period 1 of candidate slave participant 1 and time period 2 of candidate slave participant 2 is [14:00-15:00], and the intersection of time period 2 of candidate slave participant 2 and time period 3 of candidate slave participant 5 is [15:00-17:00]. The master participant determines that the computing power requirement of sample alignment algorithm 2 is not met when the number of samples of the slave participants is 2 million. In this way, the master participant can determine that candidate slave participant 1 is not the target slave participant, and the target slave participants are candidate slave participants 2 and 5 respectively. The target sample alignment algorithm is sample alignment algorithm 2, and the target time period for sample alignment is [15:00-17:00].

[0426] It can be understood that the main participant determines the target slave participant and the target sample alignment algorithm based on the second common sample alignment algorithm and also based on the time period in which the slave participant supports sample alignment and / or the number of samples supported by the slave participant for sample alignment. This can further reduce the probability of sample alignment failure or reduced efficiency due to failure to meet the computing power requirements of the sample alignment algorithm.

[0427] It should be understood that the above examples A to F are merely examples, and the master participant may also determine the target slave participant and the target sample alignment algorithm in other ways, and the embodiments of the present application do not specifically limit this.

[0428] Optionally, the first sample alignment negotiation result message is further used to indicate a target time period for sample alignment.

[0429] It can be understood that the target time period can be the intersection of the time periods indicated by the sample alignment response messages of the target slave participants, such as the target time period in Example D.

[0430] That is to say, the participants can also indicate the target time period for performing sample alignment through the first sample alignment negotiation result message, so that each participant can perform the sample alignment process within the target time period, avoiding each participant from performing the sample alignment process during the time period when other participants may not be able to meet the computing power requirements of the target sample alignment algorithm, thereby further improving the success rate and efficiency of sample alignment.

[0431] Optionally, the method flow shown in FIG6 further includes step S613.

[0432] S613: The master participant sends a second sample alignment negotiation result message to the other slave participants, excluding the target slave participant, among the multiple slave participants supporting sample alignment. Correspondingly, the other slave participants, excluding the target slave participant, among the multiple slave participants supporting sample alignment receive the second sample alignment negotiation result message from the master participant.

[0433] Among them, the second sample alignment negotiation result message is used to indicate that sample alignment is not performed, and / or the reason for not performing sample alignment. That is to say, the main participant sends the second sample alignment negotiation result message to other slave participants to indicate that sample alignment is not performed, and can promptly notify other slave participants that sample alignment will not be performed next. In addition, by indicating the reason for not performing sample alignment through the second sample alignment negotiation result message, when the main participant subsequently initiates the sample alignment negotiation process again, it can assist in adjusting the second common sample alignment algorithm indicated by the sample alignment response message fed back by the slave participant, the number of samples supported by the slave participant for sample alignment, or the time period supported by the slave participant for sample alignment, etc., so as to improve the success rate and efficiency of the sample alignment negotiation.

[0434] It can be understood that the multiple slave participants supporting the execution of sample alignment can be, for example, multiple slave participants that send a sample alignment response message from at least one candidate slave participant to indicate support for executing the sample alignment algorithm. The other slave participants among the multiple slave participants, except for the target slave participant, can, for example, refer to the slave participants among the multiple participants that the master participant has not selected as the target slave participant, such as candidate slave participant 4 in Examples A to F, candidate slave participant 2 in Example D, candidate slave participant 5 in Example E, and candidate slave participant 1 in Example F.

[0435] In addition, the reasons for not performing sample alignment may include multiple reasons. For example, for candidate slave participant 4 in Examples A to F, the reason for not performing sample alignment is that the sample alignment algorithms 3 and 4 supported by candidate slave participant 4 are not used. For another example, for candidate slave participant 2 in Example D, the reason for not performing sample alignment is that there is no intersection between time period 2 supported by candidate slave participant 2 for sample alignment and time periods 1 and 3. For another example, for candidate slave participant 5 in Example E, the reason for not performing sample alignment is that the master participant determines that the computing power requirements of sample alignment algorithm 2 are not met when the number of samples of candidate slave participant 5 is 2 million. For another example, for candidate slave participant 1 in Example F, the reason for not performing sample alignment is that there is no intersection between time period 1 and time period 3 supported by candidate slave participant 1 for sample alignment, and the master participant determines that the computing power requirements of sample alignment algorithm 2 are not met when the number of samples of candidate slave participant 1 is 2 million.

[0436] It should be understood that the above reasons for not performing sample alignment are merely examples, and in actual implementation, there may be other reasons for not performing sample alignment, which are not specifically limited in the embodiments of the present application.

[0437] In addition, the second sample alignment negotiation result message may indicate the reason for not performing the sample alignment through the parameter reason value. For details, please refer to Case 2 in the above step S611, which will not be repeated here.

[0438] It can be understood that step S613 is an optional step, that is, the master participant may not send the second sample alignment negotiation result message to the other slave participants except the target slave participant among the multiple slave participants that support the execution of sample alignment. The other slave participants can start timing when sending the sample alignment response message to the master participant. If the time threshold is exceeded and the first sample alignment negotiation result message from the master participant is not received, it can be determined that the master participant instructs not to perform sample alignment.

[0439] In addition, the above-mentioned time threshold may be pre-agreed upon by the protocol, or negotiated in advance between the candidate slave participant and the master participant, or indicated by other network elements (such as a VFL server or a master participant, etc.), and the embodiments of the present application do not specifically limit this.

[0440] It should be understood that in the above step S611, if the sample alignment response message of at least one candidate slave participant indicates that there are multiple slave participants supporting the execution of sample alignment, the master participant may determine that sample alignment cannot be performed between the master participant and the multiple slave participants based on the sample alignment response messages of the multiple slave participants, that is, the sample alignment negotiation fails. For example, there is no intersection between the second common sample alignment algorithms indicated by the sample alignment response messages of the multiple slave participants. For another example, the number (or quantity) of slave participants with an intersection between the second common sample alignment algorithms indicated by the sample alignment response messages of the multiple slave participants is less than or equal to the second threshold, that is, the minimum number of slave participants performing sample alignment. For another example, there is no intersection between the time periods supporting sample alignment indicated by the sample alignment response messages of the multiple slave participants, or the number of slave participants with an intersection is less than or equal to the second threshold, etc. The embodiments of the present application do not specifically limit this.

[0441] In addition, when the master participant determines that sample alignment cannot be performed between the master participant and the multiple slave participants, the master participant may send a second sample alignment response message to the multiple slave participants that support sample alignment to indicate that sample alignment is not performed and / or the reasons for not performing sample alignment.

[0442] It can be understood that in the event that the sample alignment negotiation fails, the main participant can initiate the sample alignment negotiation process again. For example, the main participant can adjust one or more of the first common sample alignment request, the number of samples supported by the main participant for sample alignment, and at least one time period for which the main participant supports sample alignment, and send the adjusted sample alignment request message to improve the success rate of the relevant configuration of the negotiated sample alignment.

[0443] S614: The master participant and the target slave participant execute a sample alignment process.

[0444] It is understood that based on step S612, the master participant and the target slave participant may determine a target sample alignment algorithm for performing sample alignment (e.g., may also include a target time period for sample alignment), and the master participant and the target slave participant may perform a sample alignment process based on the target sample alignment algorithm to obtain a sample alignment result. The sample alignment result may include common samples between the master participant and multiple target slave participants.

[0445] In addition, the sample alignment process may include a two-party sample alignment process and / or a multi-party sample alignment process. The specific implementation of the two-party sample alignment process and / or the multi-party sample alignment process can refer to the relevant descriptions in the preface of the specific implementation method "Sample Alignment" (such as Example 1 and Example 2), and the details will not be repeated here.

[0446] It should be understood that the master participant and the target slave participant may determine the common samples and / or common sample identifiers between the master participant and the target slave participant based on the sample alignment result.

[0447] S615: The participant sends a vertical federated learning response message to the VFL server. Correspondingly, the VFL server receives the vertical federated learning response message from the participant.

[0448] The vertical federated learning response message is used to indicate whether a participant has joined vertical federated learning. It should be understood that joining vertical federated learning can refer to joining the vertical federated learning model training process and / or the vertical federated learning inference process, which is not specifically limited in this embodiment of the application.

[0449] In addition, joining vertical federated learning and participating in vertical federated learning have the same meaning and can be expressed interchangeably. They are explained here uniformly and will not be repeated below.

[0450] It can be understood that the participants may include a master participant and a target slave participant, and the vertical federated learning response message may include a first vertical federated learning response message and a second vertical federated learning response message. Then, step S615 may include:

[0451] S615a: The target slave participant sends a first vertical federated learning response message to the VFL server. Correspondingly, the VFL server receives the first vertical federated learning response message from the target slave participant.

[0452] It can be understood that the first vertical federated learning response message can be understood as a response message to the first vertical federated learning request message in step S608, that is, the first vertical federated learning response message can also be understood as a response message of the vertical federated learning preparation process, which is used to indicate whether the target slave participant joins the vertical federated learning.

[0453] S615b: The master participant sends a second vertical federated learning response message to the VFL server. Correspondingly, the VFL server receives the second vertical federated learning response message from the master and slave participants.

[0454] It can be understood that the second vertical federated learning response message can be understood as a response message to the second vertical federated learning request message in step S609, that is, the second vertical federated learning response message can also be understood as a response message of the vertical federated learning preparation process, used to indicate whether the main participant joins the vertical federated learning.

[0455] It should be understood that the main participant and the target slave participant can determine whether to join the vertical federated learning based on the sample alignment results. For example, if there are only a small number of common samples between the first participant (the main participant and any one of the multiple target slave participants) and other participants, or the number of common samples is less than or equal to the third threshold, the first participant can determine not to join the vertical federated learning. The third threshold can be 10%, 20%, 30%, or other values ​​of the samples corresponding to the data set provided by the first participant for vertical federated learning model training, depending on the actual implementation, and the embodiments of the present application do not specifically limit this.

[0456] In addition, the first participant can also determine whether to join vertical federated learning based on other information. For example, the first participant determines that its own load is too high and decides not to participate in vertical federated learning. This embodiment of the present application does not specifically limit this.

[0457] S616. The VFL server determines the target participant to join the vertical federated learning based on the vertical federated learning response message of the participant.

[0458] It is understood that the target participants in vertical federated learning can be, for example, participants in vertical federated learning model training. After completing the sample alignment process, each target participant can identify common samples and proceed with the vertical federated learning model training process. The vertical federated learning model training process is specifically illustrated in Figure 3 and will not be further described.

[0459] Due to the information transmission method flow shown in FIG6 in the embodiment of the present application, each participant can register the sample alignment capability with the VFLSF network element. Then, when the master participant initiates vertical federated learning, the master participant can discover the VFL server, the coordinator of the vertical federated learning, through the VFLSF network element, and send a discovery request message to the VFL server, so that the VFL server discovers at least one candidate slave participant that meets the sample alignment requirements through the VFLSF network element, obtains the sample alignment capability (e.g., sample alignment algorithm) of the at least one candidate slave participant, and sends the sample alignment capability of the at least one candidate slave participant to the master participant. Then, the master participant can perform sample alignment negotiation with the at least one candidate slave participant based on the first common sample alignment algorithm jointly supported by the master participant and the at least one candidate slave participant, so as to improve the success rate and efficiency of the sample alignment negotiation. Furthermore, through the sample alignment negotiation, the master participant can determine, based on the response of the at least one candidate slave participant, that the local computing power can support the target slave participant and the target sample alignment algorithm for performing sample alignment, so as to avoid the problem of sample alignment failure caused by not meeting the computing power requirements of the sample alignment algorithm during the sample alignment process, thereby improving the success rate and efficiency of the sample alignment.

[0460] It will be appreciated that in the information transmission method flow shown in Figure 6 above, the master participant serves as the initiator of vertical federated learning. In another implementation, a slave participant can also initiate vertical federated learning, with the VFL server responsible for discovering the master participant and candidate slave participants. For example, if a slave participant discovers that training data cannot be collected due to privacy protection issues and initiates vertical federated learning, the slave participant can execute step S603 to send a discovery request message to the VFLSF network element and then determine the VFL server through step S604. Furthermore, the slave participant can execute step S605, that is, the slave participant can send a vertical federated learning initiation request message to the VFL server, so that the VFL server and the VFLSF network element execute steps S606 to S607 to determine the master participant and the candidate slave participant. After that, the VFL server sends a vertical federated learning request message to the slave participant, the master participant, and the candidate slave participant (that is, steps S608 and S609) to facilitate the execution of the sample alignment negotiation process (that is, steps S610 to S613) and the sample alignment execution process (that is, step S614) between the master participant and the slave participant, and the VFL server determines the process of the target participant to join the vertical federated learning (steps S615 to S616).

[0461] In addition, in the above example of the slave participant initiating vertical federated learning, the slave participant can initiate the sample alignment negotiation process, that is, the slave participant can execute step S610, that is, the slave participant can send a sample alignment request message to the master participant and the candidate slave participants.

[0462] It should be understood that in the information transmission method process shown in Figure 6 above, the participant (e.g., the master participant or the slave participant) can be a trusted network element in the core network (or a network element located in the same trusted domain as the VFLSF network element). For example, the participant can be an application function network element within the operator, and the participant can directly interact with network function network elements such as the VFLSF network element (or network storage function network element), the VFL server, or the network data analysis function network element. When the participant is a third-party application function network element (or called an untrusted application function (untrusted AF) network element), the participant and the above-mentioned network function network element are not in the same trusted domain, and the network open function network element can provide the participant with a network function that allows secure access to the above-mentioned network function network element, that is, the participant can interact with the above-mentioned network function network element through the network open function network element.

[0463] Below, taking the main participant in Figure 6 as an untrusted application function network element, the candidate slave participant in at least one candidate slave participant as a network data analysis function network element, the VFL server as the VFL server in Figure 5, and the VFLSF network element as the VFLSF network element or the network storage function network element in Figure 5 as an example, the information transmission method process shown in Figure 6 is further explained in combination with Figure 7.

[0464] Figure 7 is a second flow diagram of an information transmission method provided in an embodiment of the present application. As shown in Figure 7, the information transmission method flow includes: steps S701 to S728, steps S705 to S706 are the same as steps S601 to S602, steps S713 to S715 are the same as steps S606 to S608, and step S728 is the same as step S616, which will not be repeated here.

[0465] Steps S701 to S704, S707 to S712, and S716 to S727 are described below.

[0466] S701: The primary participant sends a first registration request message to a network open function network element. Correspondingly, the network open function network element receives the first registration request message from the primary participant.

[0467] It can be understood that the first registration request message may be specifically referred to the registration request message in step S601 of FIG6 , which will not be described in detail here.

[0468] It should be understood that the network open function network element can support the following functions: capability and event exposure, verification and authorization, and internal and external information conversion. Among them, capability and event exposure can, for example, mean that a network function network element within the network can securely disclose capabilities and events to an untrusted application function network element through the network open function network element. Verification and authorization can, for example, mean that the network open function network element can verify and authorize the behavior of the untrusted application function network element to enable the untrusted application function network element to securely send information to the network function network element. Internal and external information conversion can, for example, mean: converting the information exchanged between the untrusted application function network element and the network function network element. In addition, to improve security, the network open function network element can also shield the information of the untrusted application function network element and the sensitive information of the network function network element.

[0469] Exemplarily, the primary participant may send a first registration request message by invoking a first service operation provided by a network open function element. The first service operation may be a service operation related to vertical federated learning registration. For example, the first service operation may be a vertical federated learning registration request service operation.

[0470] In addition, the main participant can interact with the network open function network element through the service interface Nnef of the network open function network element.

[0471] It should be understood that the above-mentioned main participant sending the first registration request message by calling the first service operation provided by the network open function network element is only an example. The main participant can also send the first registration request message by calling other service functions or service operations provided by the network open function network element, or use other methods to send the first registration request message. The embodiments of the present application do not make specific limitations on this.

[0472] It can be understood that the network open function network element can verify and authorize the first registration request message of the primary participant, and then if the authorization is passed, the network open function network element can provide the information contained in the first registration request message to the VFLSF network element (ie, step S702).

[0473] In addition, the execution subject of step S701 being the master participant is merely an example. As shown in the relevant description of step S601 in FIG6 , the participants in the vertical federated learning in the embodiment of the present application may not distinguish between the master participant and the slave participant, that is, the execution subject of step S701 may be an untrusted network element, and the master participant and the slave participant may not be distinguished. In addition, as shown in the relevant description of step S605 in FIG6 , when the participants distinguish between the master participant and the slave participant, the initiator of the vertical federated learning may also be a slave participant, that is, the execution subject of step S701 may also be a slave participant. This is uniformly described here and will not be repeated below.

[0474] S702: The network open function network element sends a second registration request message to the VFLSF network element. Correspondingly, the VFLSF network element receives the second registration request message from the network open function network element.

[0475] It can be understood that the second registration request message may include the information included in the first registration request message.

[0476] Exemplarily, the network open function network element may send the second registration request message by invoking a second service operation provided by the VFLSF network element. If the VFLSF network element is the VFLSF network element in FIG5 , the second service operation may be a service operation related to vertical federated learning registration provided by the VFLSF network element (e.g., Nvflsf Management Register Request). If the VFLSF network element is the network storage function network element in FIG5 , the second service operation may be a network function management service operation provided by the network storage function network element (e.g., Nnrf_NFManagement NFRegister Request).

[0477] In addition, the above-mentioned second service operation may be different from the first service operation in step S701, that is, the main participant sends a first registration request message to the network open function network element, and the network open function network element sends a second registration request message to the VFLSF network element, and the service operations used between the two are different.

[0478] It should be understood that the above-mentioned network open function network element sending the second registration request message by calling the second service operation provided by the VFLSF network element is only an example. The network open function network element can also send the second registration request message by calling other service functions or service operations provided by the VFLSF network element, or send the second registration request message in other ways. The embodiments of the present application do not make specific limitations on this.

[0479] S703: The VFLSF network element sends a second registration response message to the network openness function network element. Correspondingly, the network openness function network element receives the second registration response message from the VFLSF network element.

[0480] It can be understood that the second registration response message is used to indicate whether the registration is successful.

[0481] In addition, the VFLSF network element may send the second registration response message by invoking a third service operation. For example, if the VFLSF network element is the VFLSF network element in FIG5 , the third service operation may be a service operation related to vertical federation registration provided by the VFLSF network element (e.g., Nvflsf Management Register Response). For another example, if the VFLSF network element is a network storage function network element, the third service operation may be a network function management service operation provided by the network storage function network element (e.g., Nnrf_NFManagement NFRegister Response).

[0482] S704: The network open function network element sends a first registration response message to the primary participant. Correspondingly, the primary participant receives the first registration response message from the network open function network element.

[0483] It can be understood that the information included in the first registration response message may include the information included in the second registration response message.

[0484] In addition, the network open function network element may send the first registration response message by calling a fourth service operation. For example, the fourth service operation may be a vertical federated learning registration response service operation provided by the network open function network element.

[0485] S707: The active participant sends a first discovery request message to the network open function network element. Correspondingly, the network open function network element receives the first discovery request message from the active participant.

[0486] It can be understood that the first discovery request message can be specifically referred to the discovery request message in step S603 of FIG6 , which will not be described in detail here.

[0487] In addition, the primary participant may send the first discovery request message by calling a related service operation provided by the network open function network element. For example, the primary participant may send the first discovery request message by calling the vertical federated learning participant discovery request (Nnef VFL ParticipantDiscovery Request) service operation provided by the network open function.

[0488] It can be understood that when the network open function network element verifies and authorizes the first discovery request message sent by the primary participant, and then if the authorization is passed, the network open function network element can provide the VFLSF network element with the information contained in the first discovery request message (ie step S708).

[0489] S708: The network open function network element sends a second discovery request message to the VFLSF network element. Correspondingly, the VFLSF network element receives the second discovery request message from the network open function network element.

[0490] It can be understood that the second discovery request message may include the information included in the first discovery request message.

[0491] Exemplarily, if the VFLSF network element is the VFLSF network element in FIG5 , the network open function network element may send the second discovery request message by calling the vertical federated learning participant discovery request (Nvflsf VFL ParticipantDiscovery Request) service operation provided by the VFLSF network element. If the VFLSF network element is the network storage function network element in FIG5 , the network open function network element may send the second discovery request message by calling the network function discovery request (Nnrf NFDiscovery Request) service operation provided by the network storage function network element.

[0492] S709: The VFLSF network element sends a second discovery response message to the network openness function network element. Correspondingly, the network openness function network element receives the second discovery response message from the VFLSF network element.

[0493] It can be understood that the second discovery response message can be specifically referred to the discovery response message in step S604 of FIG6 , which will not be described in detail here.

[0494] S710: The network open function network element sends a first discovery response message to the primary participant. Correspondingly, the primary participant receives the first discovery response message from the network open function network element.

[0495] In a possible implementation, the first discovery response message includes the information included in the second discovery response message. In other words, the first discovery response message may include the identifier or address of at least one VFL server included in the discovery response message in step S604.

[0496] In another possible implementation, the first discovery response message includes a first identifier or a first address assigned (or configured, provided, etc.) by the network open function network element to at least one VFL server indicated by the second discovery response message, the first identifier of each VFL server in the at least one VFL server being different from the identifier of each VFL server, and the first address of each VFL server being different from the address of each VFL server. In other words, the network open function network element can assign a first identifier or a first address to each VFL server in the at least one VFL server indicated by the second discovery response message to avoid exposing the true identifier or true address of the VFL server, thereby improving security.

[0497] For example, assume that the at least one VFL server indicated by the second discovery response message includes: VFL server 1, VFL server 2, and VFL server 3, the identifier of VFL server 1 is identifier A, the identifier of VFL server 2 is identifier B, and the identifier of VFL server 3 is identifier C. Based on the above conditions, the network open function network element allocates a first identifier of identifier #1 to VFL server 1, a first identifier of identifier #2 to VFL server 2, and a first identifier of identifier #3 to VFL server 3. The first discovery response message includes identifier #1, identifier #2, and identifier #3, so that the primary participant obtains the first identifier of the VFL server instead of the true identifier of the VFL server, thereby avoiding exposure of the true identifier of the VFL server.

[0498] In addition, the principle of allocating the first address to the at least one VFL server by the network open function network element is similar to the above-mentioned example of allocating the first identifier, which is not repeated here.

[0499] It can be understood that the network open function network element can save or maintain the correspondence between the identifier of the VFL server and the first identifier of the VFL server, and / or the correspondence between the address of the VFL server and the first address of the VFL server. In this way, the network open function network element can determine the real identifier of the VFL server according to the first identifier of the VFL server based on the above correspondence, and / or determine the real address of the VFL server according to the first address of the VFL server.

[0500] It can be understood that the above description of the first identifier is only an example, and the network open function network element can also use other methods to allocate the first identifier, which is not specifically limited in the embodiments of the present application.

[0501] S711: The primary participant sends a first vertical federated learning initiation request message to the network open function network element. Correspondingly, the network open function network element receives the first vertical federated learning initiation request message from the primary participant.

[0502] It can be understood that the first vertical federated learning initiation request message can be specifically referred to the vertical federated learning initiation request message in step S605 of Figure 6, and will not be repeated here.

[0503] In addition, as described in the relevant description of step S605, the main participant can determine a VFL server from at least one VFL server, and then the first vertical federated learning initiation request message in step S711 may include the identifier and / or address of the VFL server. In this way, the network open function network element can determine which VFL server among the at least one VFL server the information contained in the first vertical federated learning initiation request message is sent to based on the identifier and / or address of the VFL server.

[0504] It can be understood that when the first discovery response message in step S710 includes the first identifier and / or first address of each VFL server in at least one VFL server, the first vertical federated learning initiation request message includes the first identifier and / or first address of the VFL server determined by the main participant from at least one VFL server.

[0505] It should be understood that the VFL server indicated by the first vertical federated learning initiation request message in the above step S711 corresponds to the primary participant, that is, the VFL server is the VFL server corresponding to the vertical federated learning process initiated by the primary participant. Since the network open function network element may provide services to initiators of other vertical federated learning processes other than the primary participant, in order to distinguish the VFL server corresponding to the primary participant from the VFL servers corresponding to other initiators, the network open function network element can maintain the corresponding VFL servers between different initiators. In other words, by maintaining the correspondence between the primary participant and the VFL server selected by the primary participant, the network open function network element can determine to which of the multiple VFL servers the message of the primary participant is to be sent, and to which of the multiple initiators the message of the VFL server is to be sent, based on the correspondence.

[0506] In one possible implementation, the first vertical federated learning initiation request message also includes a correlation ID, which associates the vertical federated learning process between the master participant and the VFL server. It can be understood that the master participant may execute multiple different vertical federated learning processes in parallel, and the multiple different vertical federated learning processes can be distinguished by the correlation ID. For example, taking the vertical federated learning process as a sample alignment process, assuming that the master participant and slave participant 1 execute sample alignment process #1, and the master participant and slave participant 2 execute sample alignment process #2, the two sample alignment processes are different, for example, the common samples determined by the two sample alignment processes are used for two different vertical federated learning model training, and for example, the sample alignment algorithms and / or data sets used for vertical federated learning model training are different in the two sample alignment processes. In addition, the vertical federated learning process can also be a vertical federated learning model training process or a vertical federated learning reasoning process. The master participant can execute multiple different vertical federated learning model training processes or vertical federated learning reasoning processes in parallel with different slave participants.

[0507] Furthermore, the multiple different vertical federated learning processes executed in parallel by the above-mentioned master participant may correspond to the same VFL server, and then by adding the association identifier in the message exchanged between the master participant, the VFL server, or the slave participant through the network open function network element, the master participant, the VFL server, the slave participant, or the network open function network element can distinguish the multiple different vertical federated learning processes executed in parallel. For example, the network open function network element can determine which vertical federated learning process of the master participant the first vertical federated learning initiation request message belongs to based on the association identifier, and when the network open function network element sends the information contained in the first vertical federated learning initiation request message to the VFL server corresponding to the vertical federated learning process of the master participant, it can also send the association identifier, so that the VFL server can determine which vertical federated learning process the received message is for based on the association identifier.

[0508] It can be understood that the above-mentioned association identifier is allocated (or generated) by the main participant. In another implementation, the network open function network element can also allocate the association identifier. For details, please refer to the following step S712.

[0509] It should be understood that the above examples of the vertical federated learning process are only a few limited examples. The vertical federated learning process can also include a vertical federated learning reasoning process, a sample alignment process, etc., and the embodiments of this application do not make specific limitations on this.

[0510] S712: The network open function network element sends a second vertical federated learning initiation request message to the VFL server. Correspondingly, the VFL server receives the second vertical federated learning initiation request message from the network open function network element.

[0511] It can be understood that the VFL server may be the VFL server in step S605, and the second vertical federated learning initiation request message includes the information included in the first vertical federated learning initiation request message.

[0512] In one possible implementation, the second vertical federated learning initiation request message may include a second identifier and / or a second address allocated by the network open function element to the primary participant. It is understood that, similar to the allocation of a first identifier and / or a first address by the network open function element to at least one VFL server in step S710, the network open function element may also allocate a second identifier and / or a second address to the primary participant to avoid exposing the primary participant's true identifier and true address.

[0513] In addition, in the following steps, when the network open function network element sends a message from the main participant to other network elements, the network open function network element can add the second identifier and / or second address of the main participant in the message. This is explained here uniformly and will not be repeated below.

[0514] It can be understood that when the first vertical federated learning initiation request message does not include an association identifier (that is, when the primary participant has not assigned an association identifier), the network open function network element can assign the association identifier and carry the association identifier in the second vertical federated learning initiation request message. For example, the network open function network element can determine the vertical federated learning process #1 of the primary participant corresponding to the first vertical federated learning initiation request message based on the information contained in the first vertical federated learning initiation request message, and assign an association identifier to the vertical federated learning process #1. In addition, the network open function network element can determine the VFL server selected by the primary participant based on the first vertical federated learning initiation request message, and add the association identifier in the second vertical federated learning initiation request message sent to the VFL server. In this way, the VFL server can determine that the second vertical federated learning initiation request message belongs to the vertical federated learning process #1 based on the association identifier. In addition, after the network open function network element allocates the association identifier, it can send the association identifier to each participant of the vertical federated learning process corresponding to the association identifier (for example, the main participant, the VFL server, or at least one candidate slave participant, etc.), so that each participant of the vertical federated learning process can determine which vertical federated learning process the sent or received message belongs to based on the association identifier. For example, for the VFL server, the network open function network element can send the association identifier to the VFL server through the second vertical federated learning initiation request message, and then the VFL server can determine that the second vertical federated learning initiation request message is a request message of the vertical federated learning process from the main participant. In addition, the VFL server executes steps S713 to S714 (i.e., the VFL server obtains information of at least one candidate slave participant) based on the second vertical federation initiation request message, and then triggers the VFL server to send a second vertical federation learning request message to the network open function network element (i.e., step S716). Therefore, when the network open function network element receives the second vertical federation learning request message from the VFL server and triggers the sending of a third vertical federation learning request message to the master participant (i.e., step S717), it can send an association identifier to the master participant through the third vertical federation learning request message.

[0515] For another example, in the subsequent sample alignment negotiation process, the network open function network element may send an association identifier to at least one candidate slave participant through the second sample alignment request message in step S719.

[0516] S716: The VFL server sends a second vertical federated learning request message to the network open function network element. Correspondingly, the network open function network element receives the second vertical federated learning request message from the VFL server.

[0517] It can be understood that the second vertical federated learning request message can be specifically referred to the second vertical federated learning request message in step S609 of Figure 6, and will not be repeated here.

[0518] In addition, according to the relevant instructions on the association identifier in step S711 and step S712, the second vertical federated learning request message may also include the association identifier corresponding to the vertical federated learning process between the main participant, the VFL server, and at least one candidate slave participant indicated by the second vertical federated learning request message, so as to facilitate the network open function network element to determine that the second vertical federated learning request message belongs to the vertical federated learning process corresponding to the association identifier.

[0519] It can be understood that, based on the network open function network element maintaining the correspondence between the main participant and the VFL server in the aforementioned step S711, the network open function network element can also maintain the correspondence between the main participant, the VFL server, and at least one candidate slave participant indicated by the second vertical federated learning request message, and then when the main participant (untrusted application function network element) interacts with at least one candidate slave participant (for example, the network data analysis function network element) through the network open function network element during the vertical federated learning process, the network open function network element can determine to which network analysis function network element the main participant's message is sent, and to which application function network element the network data analysis function network element's message is sent based on the correspondence.

[0520] S717: The network open function network element sends a third vertical federated learning request message to the primary participant. Correspondingly, the primary participant receives the third vertical federated learning request message from the network open function network element.

[0521] In one possible implementation, the third vertical federated learning request message includes a third identifier and / or a third address assigned by the network open function element to at least one candidate slave participant. It is understood that, similar to the assignment of the first identifier and / or first address by the network open function element to at least one VFL server in step S710, the third identifier and / or third address assigned by the network open function element can avoid exposing the true identifier and true address of at least one candidate slave participant, thereby improving security.

[0522] For example, assuming that at least one candidate slave participant includes network data analysis function network element 1 and network data analysis function network element 2, the second identifier assigned by the network open function network element to the master participant is ID#1, the third identifier assigned to the network data analysis function network element 1 is ID#2, and the third identifier assigned to the network data analysis function network element 2 is ID#3. When the master participant sends a message to the network data analysis function network element 1 through the network open function network element (such as the first sample alignment request message in step S718 below), the network open function network element carries ID#1 when sending the message to the network data analysis function network element 1, so that the network data analysis function network element 1 determines which participant sent the message based on ID#1. When the network data analysis function network element 2 sends a message to the master participant through the network open function network element, the network open function network element carries ID#3 when sending the message to the master participant, so that the master participant determines that the message was sent by the network data analysis function network element 2 based on ID#3.

[0523] It can be understood that according to the above steps S711, S712, and the description of the association identifier in the above step S716, the third vertical federated learning request message can also include the association identifier in step S716, so as to distinguish multiple different vertical federated learning processes executed in parallel by the main participants. For details, please refer to the relevant description of step S711, which will not be repeated here.

[0524] In addition, in the following steps involving the interaction between the master participant and at least one candidate slave participant (for example, the steps in the sample alignment negotiation process), the messages interacting between the master participant and at least one candidate slave participant (including the target slave participant) may carry the association identifier, third identifier and / or third address in the above-mentioned step S716, which are uniformly explained here and will not be repeated below.

[0525] The sample alignment negotiation process is introduced below. The sample alignment negotiation process includes steps S718 to S724.

[0526] S718: The primary participant sends a first sample alignment request message to the network open function network element. Correspondingly, the network open function network element receives the first sample alignment request message from the primary participant.

[0527] It can be understood that the first sample alignment request message may specifically refer to the sample alignment request message in step S610 of FIG6 , which will not be described in detail here.

[0528] In addition, the first sample alignment request message may further include the association identifier in step S717, so that the network open function network element can determine to which vertical federated learning process of the primary participant the first sample alignment request message belongs.

[0529] In addition, the first sample alignment request message may also include a third identifier and / or a third address of each candidate slave participant in at least one candidate slave participant in step S717, so that the network open function network element can determine which network function network elements (such as a network data analysis function network element) among the multiple network function network elements in the network to send the information contained in the first sample alignment request message.

[0530] S719: The network open function network element sends a second sample alignment request message to at least one candidate slave participant. Correspondingly, the at least one candidate slave participant receives the second sample alignment request message from the network open function network element.

[0531] It can be understood that the second sample alignment request message includes the information included in the first sample alignment request message.

[0532] In addition, the second sample alignment request message may include the second identifier and / or second address of the master participant, so that at least one candidate slave participant can determine that the second sample alignment request message is requested by the master participant.

[0533] In addition, considering that at least some of the at least one candidate slave participants may also participate in multiple different vertical federated learning processes in parallel, the second sample alignment request message may also include the association identifier in step S718, so that at least one candidate slave participant can determine the multiple different vertical federated learning processes in which it participates in parallel through the association identifier.

[0534] S720: At least one candidate slave participant sends a second sample alignment response message to the network openness function network element. Correspondingly, the network openness function network element receives the second sample alignment response message from the at least one candidate slave participant.

[0535] It can be understood that the second sample alignment response message may specifically refer to the sample alignment response message in step S611 of FIG6 , which will not be described in detail here.

[0536] In addition, the second sample alignment response message may further include the association identifier in step S719, so that the network open function network element can distinguish which vertical federated learning process the second sample alignment response message belongs to.

[0537] S721: The network open function network element sends a first sample alignment response message to the primary participant. Correspondingly, the primary participant receives the first sample alignment response message from the network open function network element.

[0538] It can be understood that the first sample alignment response message includes the information included in the second sample alignment response message.

[0539] In addition, the first sample alignment response message may include a third identifier and / or a third address of at least one candidate slave participant, so that the master participant can determine which candidate slave participant responds to the first sample alignment response message.

[0540] In addition, the first sample alignment response message may further include the association identifier in step S720 , so that the primary participant can distinguish which vertical federated learning process the first sample alignment response message belongs to.

[0541] S722: The active participant sends a first sample alignment negotiation result message to the network openness function network element. Correspondingly, the network openness function network element receives the first sample alignment negotiation result message from the active participant.

[0542] It can be understood that the first sample alignment negotiation result message can be specifically referred to the first sample alignment negotiation result message in step S612 of FIG6 , and will not be described in detail here.

[0543] In addition, the first sample alignment negotiation result message may further include the association identifier in step S721, or the third identifier and / or third address of the target slave participant, etc., which is not specifically limited in this embodiment of the present application.

[0544] S723: The network open function network element sends a second sample alignment negotiation result message to the target slave participant. Correspondingly, the target slave participant receives the second sample alignment negotiation result message from the network open function network element.

[0545] It can be understood that the second sample alignment negotiation result message includes the information included in the first sample alignment negotiation result message.

[0546] In addition, the second sample alignment negotiation result message may further include the second identifier and / or second address of the primary participant, or the association identifier in step S722, etc., which is not specifically limited in this embodiment of the present application.

[0547] Optionally, the information transmission method process shown in FIG7 may further include steps S724 and S725.

[0548] S724: The active participant sends a third sample alignment negotiation result message to the network openness function network element. Correspondingly, the network openness function network element receives the third sample alignment negotiation result message from the active participant.

[0549] It can be understood that the third sample alignment negotiation result message can be specifically referred to the second sample alignment negotiation result message in step S613 of FIG6 , and will not be described in detail here.

[0550] In addition, the third sample alignment negotiation result message may also include a third identifier and / or a third address, or an associated identifier, etc. of at least one candidate slave participant other than the target slave participant, which is not specifically limited in this embodiment of the present application.

[0551] S725. The network open function network element sends a fourth sample alignment negotiation result message to the slave participants other than the target slave participant among the multiple slave participants supporting sample alignment. Correspondingly, the slave participants other than the target slave participant among the multiple slave participants supporting sample alignment receive the fourth sample alignment negotiation result message from the network open function network element.

[0552] It can be understood that the fourth sample alignment negotiation result message includes the information included in the third sample alignment negotiation result message.

[0553] In addition, the fourth sample alignment negotiation result message may further include the second identifier and / or second address of the main participant, or an associated identifier, etc., which is not specifically limited in this embodiment of the present application.

[0554] S726: The master participant, the network open function network element, and the target slave participant perform a sample alignment process.

[0555] It is understood that step S726 is similar to step S614, differing in that the master participant is an untrusted application function element, and the target slave participant is a network data analysis function element. Consequently, information exchange during the sample alignment process between the master and target slave participants is forwarded and coordinated by the network open function element. In other words, the network open function element is responsible for maintaining the correspondence between the master participant and one or more target slave participants, as well as distinguishing between the vertical federated learning process. For details, please refer to the relevant description of step S716 and will not be repeated here.

[0556] In addition, the messages exchanged between the main participant, the network open function network element, and the target slave participant may carry the association identifier, second identifier, third identifier, second address, or third address involved in the above steps, etc., and the embodiments of the present application do not make specific limitations on this.

[0557] S727: The participant sends a vertical federated learning response message to the VFL server. In response, the VFL server receives the vertical federated learning response message from the participant.

[0558] It can be understood that step S727 is similar to step S615 of FIG6 . In the case where the participants include a master participant and a target slave participant, step S727 may include:

[0559] S727a: The target slave participant sends a first vertical federated learning response message to the VFL server. Correspondingly, the VFL server receives the first vertical federated learning response message from the target slave participant.

[0560] It can be understood that the specific implementation of step S727a can refer to step S615a, which will not be repeated here.

[0561] S727b: The primary participant sends a second vertical federated learning response message to the network open function network element. Correspondingly, the network open function network element receives the second vertical federated learning response message from the primary participant.

[0562] It can be understood that the second vertical federated learning response message can be specifically referred to the second vertical federated learning response message in step S615b of Figure 6, and will not be repeated here.

[0563] In addition, the second vertical federated learning response message may also include the first identifier and / or first address of the VFL server, or an associated identifier, etc., which is not specifically limited in this embodiment of the present application.

[0564] S727c: The network open function network element sends a third vertical federated learning response message to the VFL server. Correspondingly, the VFL server receives the third vertical federated learning response message from the network open function network element.

[0565] It can be understood that the third vertical federated learning response message includes the information contained in the second vertical federated learning response message.

[0566] In addition, the third vertical federated learning response message may include the second identifier and / or second address of the primary participant, or an associated identifier, etc., which is not specifically limited in this embodiment of the present application.

[0567] Due to the information transmission method process shown in Figure 7 in the embodiment of the present application, each participant can register the sample alignment capability with the VFLSF network element. The master participant can discover the VFL server, the coordinator of the vertical federated learning, through the VFLSF network element, and send a discovery request message through the VFL server to obtain at least one candidate slave participant and the sample alignment capability (for example, the sample alignment algorithm) of the at least one candidate slave participant. Then, the master participant can conduct sample alignment negotiation with the at least one candidate slave participant based on the first common sample alignment algorithm jointly supported by the master participant and the at least one candidate slave participant to improve the success rate and efficiency of the sample alignment negotiation. Moreover, through the sample alignment negotiation, the master participant can determine, based on the response of at least one candidate slave participant, that the local computing power can support the target slave participant and the target sample alignment algorithm for executing the sample alignment, so as to avoid the problem of sample alignment failure caused by failure to meet the computing power requirements of the sample alignment algorithm during the sample alignment process, thereby improving the success rate and efficiency of the sample alignment. Furthermore, the above-mentioned main participant can be an untrusted network element, and the network open function network element can provide coordination, identity allocation, and trusted interaction functions for the interaction between the main participant and network elements in other trusted domains, so as to better realize the interaction between the above-mentioned main participant, at least one candidate slave participant, VFL server, and VFLSF network element.

[0568] It can be understood that, as described in FIG5 regarding the VFLSF network element, the VFLSF network element can be a network open function network element, or the network open function network element has the function of the VFLSF network element, so that the main participant in FIG7 (i.e., the untrusted application function network element) can interact directly with the VFLSF network element. For example, for steps S701 to S704, the main participant can directly send a first registration request message to the VFLSF network element, and receive a first registration response message from the VFLSF network element. For steps S707 to S710, the main participant can directly send a first discovery request message to the VFLSF network element, and receive a first discovery response message from the VFLSF network element, and the VFLSF network element can assign a first identifier and / or a first address to each VFL server in at least one VFL server. For details, please refer to step S710, which will not be repeated here.

[0569] In addition, for steps S711 to S712, and subsequent steps S716 to S728, the VFLSF network element replaces the network open function network element, that is, the VFLSF network element maintains the correspondence between the master participant, the VFL server, and at least one candidate slave participant, the association identifier corresponding to the vertical federated learning process between the master participant, the VFL server, and at least one candidate slave participant, assigns a second identifier and / or a second address to the master participant, and assigns a third identifier and / or a third address to at least one candidate slave participant, etc. For details, please refer to the relevant instructions in Figure 7, which will not be repeated here.

[0570] It should be understood that in the information transmission method processes shown in Figures 6 and 7 above, the master participant can also replace the VFL server to be responsible for coordination, that is, to discover candidate slave participants. This will be explained below in conjunction with the information transmission method process shown in Figure 8.

[0571] Figure 8 is a third flow diagram of an information transmission method provided in an embodiment of the present application. As shown in Figure 8, the flow of the information transmission method includes: steps S801 to S812, steps S801 to S804 are the same as steps S601 to S604, steps S807 to S810 are the same as steps S610 to S613, and step S811 is the same as step S614, and will not be repeated here.

[0572] It should be understood that the above steps S803, S804, and S812 are optional steps, that is, the VFL server may not participate in the sample alignment and longitudinal federated learning model training process.

[0573] Steps S805, S806, and S812 are described below respectively.

[0574] S805. The primary participant sends a participant discovery request message to the VFLSF network element. Accordingly, the VFLSF network element receives the participant discovery request message from the primary participant. The participant discovery request message is used to request the discovery of participants. The participant discovery request message may include: a vertical federated learning group identifier, a vertical federated learning model type identifier, slave participant requirements, a second vertical federated learning capability type indication, and a sample alignment requirement. For details, see step S606 in Figure 6, which will not be repeated here.

[0575] It can be understood that when the primary participant is an untrusted application function network element, and the VFLSF network element is the VFLSF network element or the network storage function network element in Figure 5, the primary participant sends a participant discovery request message to the VFLSF network element, which is similar to the information transmission process shown in Figure 7. The primary participant can send a participant discovery request message to the VFLSF network element through the network open function network element. For example, the primary participant sends a first participant discovery request message to the network open function network element, and the network open function network element verifies and authorizes the primary participant's request. If the authorization is passed, the network open function network element sends a second participant discovery request message to the VFLSF network element. The second participant discovery request message includes the content contained in the first participant discovery request message.

[0576] In addition, similar to step S707 of Figure 7, the main participant can send the first participant discovery request message by calling the Nnef VFL ParticipantDiscovery Request service operation of the network open function network element, and the network open function network element can send the second participant discovery request message by calling the Nnrf NFDiscovery Request service operation of the network storage function network element.

[0577] It can also be understood that when the primary participant is an untrusted application function network element and the VFLSF network element is a network open function network element, the primary participant can execute step S805, ie, directly send a participant discovery request message to the VFLSF network element.

[0578] S806. The VFLSF network element sends a participant discovery response message to the master participant. Accordingly, the master participant receives the participant discovery response message from the VFLSF network element. The participant discovery response message includes information about at least one candidate slave participant. The information about the at least one candidate slave participant may include an identifier or address of the at least one candidate slave participant. The identifier or address of the at least one candidate slave participant can be used by the master participant to communicate with the at least one candidate slave participant.

[0579] It can be understood that the participant discovery response message in step S806 can be specifically referred to step S607 in Figure 6, and will not be repeated here.

[0580] It should be understood that when the primary participant is an untrusted application function network element and the VFLSF network element is the VFLSF network element or the network storage function network element in Figure 5, the VFLSF network element sends a participant discovery response message to the primary participant in a similar manner to the information transmission process shown in Figure 7. The VFLSF network element can send a participant discovery response message to the primary participant via the network open function network element. For example, the VFLSF network element sends a second participant discovery response message to the network open function network element, and the network open function network element sends a first participant discovery response message to the primary participant, where the first participant discovery response message includes the information contained in the second participant discovery response message.

[0581] It can be understood that, similar to the network open network element in Figure 7, the network open function network element maintains the correspondence between the main participant and at least one candidate slave participant, and then when the main participant (untrusted application function network element) interacts with at least one candidate slave participant (such as the network data analysis function network element) through the network open function network element in the vertical federated learning process (such as the sample alignment process, the vertical federated learning model training process, or the vertical federated learning inference process), the network open function network element can determine to which network analysis function network element the main participant's message is sent, and to which application function network element the network data analysis function network element's message is sent based on the correspondence.

[0582] In addition, the network open function network element can assign an association identifier to the vertical federated learning process between the master participant and at least one candidate slave participant to distinguish multiple vertical federated learning processes executed in parallel by the master participant. For details, please refer to steps S711 and S712, which will not be repeated here.

[0583] In addition, the network open function network element can also assign a second identifier and / or a second address to the main participant, and assign a first identifier and / or a first address to at least one slave participant, thereby avoiding exposing the true identifier and true address of the main participant, and avoiding exposing the true identifier and true address of at least one candidate slave participant. For details, please refer to the relevant instructions of step S712 and step S717, which will not be repeated here.

[0584] It should be understood that in the following steps involving the interaction between the master participant and at least one candidate slave participant, the messages interacting between the master participant and at least one candidate slave participant (including the target slave participant) may carry the association identifier, third identifier and / or third address corresponding to the vertical federated learning process between the above-mentioned master participant and at least one candidate slave participant. These are explained here uniformly and will not be repeated below.

[0585] S812. The main participant sends a vertical federated learning model training request message to the VFL server. Correspondingly, the VFL server receives the vertical federated learning model training request message from the main participant. The vertical federated learning model training request message is used to indicate the target participant who joins the vertical federated learning. It can be understood that the target participant who joins the vertical federated learning can refer to the participant who joins the vertical federated learning model training, and then after completing the sample alignment process (i.e., step S811), each target participant can determine the common sample, and then the vertical federated learning model training process can be carried out. In the vertical federated learning model training process, the VFL server can be responsible for the decryption and transmission of the intermediate results between the participants. For details, please refer to Figure 3, which will not be repeated here.

[0586] In addition, the main participant can determine the target participant participating in the vertical federated learning based on the sample alignment result obtained by executing the sample alignment process in step S811. For details, please refer to step S615b, which will not be repeated here.

[0587] Due to the information transmission method process shown in Figure 8 in the embodiment of the present application, each participant can register the sample alignment capability with the VFLSF network element, and then when the master participant initiates vertical federated learning, the master participant can discover at least one candidate slave participant that meets the sample alignment requirements through the VFLSF network element, and obtain the sample alignment capability (such as the sample alignment algorithm) of the at least one candidate slave participant. Then, the master participant can conduct sample alignment negotiation with the at least one candidate slave participant based on the first common sample alignment algorithm jointly supported by the master participant and the at least one candidate slave participant, so as to improve the success rate and efficiency of the sample alignment negotiation. Furthermore, through the sample alignment negotiation, the master participant can determine that the local computing power can support the target slave participant and the target sample alignment algorithm for executing the sample alignment based on the response of at least one candidate slave participant, so as to avoid the problem of sample alignment failure caused by failure to meet the computing power requirements of the sample alignment algorithm during the sample alignment process, thereby improving the success rate and efficiency of the sample alignment.

[0588] It should be understood that in the information transmission method process shown in Figure 8 above, the master participant serves as the initiator of vertical federated learning. In another implementation, the slave participant can also initiate vertical federated learning. The slave participant is responsible for discovering the master participant, and then the master participant is responsible for discovering at least one candidate slave participant. For example, the slave participant can discover the master participant through the VFLSF network element, and then notify the master participant to discover at least one candidate slave participant (i.e., the master participant executes steps S805 to S806). The master participant can then execute steps S808 to S812 in Figure 8.

[0589] In another implementation, the slave participant may also initiate vertical federated learning, and the slave participant is responsible for discovering the master participant and at least one candidate slave participant. For example, the slave participant may discover the master participant and at least one candidate slave participant through the VFLSF network element, and the parameters in the request message of step S805 also include the parameters for discovering the master participant, that is, the request message in step S805 may include a combination of parameters corresponding to discovering the master participant and discovering the slave participant. It can be understood that in the case where vertical federated learning does not distinguish between master and slave participants, step S805 includes parameters corresponding to discovering the participant (for example, discovering the VFL client).

[0590] In addition, for the above step S805, the master participant corresponding to the untrusted application function network element, the VFLSF network element is the relevant description of the VFLSF network element or the network storage function network element in Figure 5, and the first participant discovery request message can be used to request the discovery of the master participant and the slave participant; or, without distinguishing between the master and slave participants, the first participant discovery request can be used to request the discovery of the participant (i.e., the VFL client).

[0591] Figure 9 is a fourth flow diagram of an information transmission method provided in an embodiment of the present application. This information transmission method is applicable to interactions between a first network element and a second network element. The first network element, through a first request, indicates at least one first sample alignment algorithm to the second network element, triggering the second network element to determine whether to support sample alignment using the at least one first sample alignment algorithm and to provide a first response back to the first network element. This facilitates negotiation of the sample alignment algorithm, avoids sample alignment failures due to unsatisfied computing power requirements of the sample alignment algorithm during sample alignment, and improves the success rate and efficiency of sample alignment.

[0592] As shown in FIG9 , the process of the information transmission method includes:

[0593] S901: A first network element sends a first request to a second network element. Correspondingly, the second network element receives the first request from the first network element. The first request includes first information, and the first information is used to indicate at least one first sample alignment algorithm supported by the first network element.

[0594] S902: The second network element sends a first response to the first network element. Correspondingly, the first network element receives the first response from the second network element.

[0595] The first response is used to indicate support for sample alignment, and / or a second sample alignment algorithm in at least one first sample alignment algorithm, the second sample alignment algorithm being used to determine common samples between a data set of the first network element for training a longitudinal federated learning model and a data set of the second network element for training a longitudinal federated learning model; or, the first response is used to indicate support for sample alignment, and / or the reason for not supporting sample alignment.

[0596] Steps S901 and S902 are described below respectively.

[0597] For step S901:

[0598] It can be understood that the first network element is a participant in vertical federated learning. For example, the first network element is the master participant in vertical federated learning model training, and the second network element is a candidate slave participant in vertical federated learning model training. For another example, the first network element is a slave participant in vertical federated learning model training, and the second network element is a candidate slave participant or master participant in vertical federated learning model training. That is, in the embodiment of the present application, either the master participant or the slave participant can initiate the sample alignment negotiation process, i.e., send the first request.

[0599] It should be understood that without distinguishing between master and slave participants, the first network element and the second network element are both participants in the vertical federated learning model training.

[0600] Furthermore, since the first network element is a participant in the vertical federated learning model training, and the participant can be a network function network element or a terminal device, the first network element can be a network function network element, a terminal device, or an access network device. For example, the first network element can be the network data analysis function network element, application function network element, access and mobility management function network element, session management function network element, or policy control function network element in Figure 5. Furthermore, the second network element is similar to the first network element and can be a network function network element, a terminal device, or an access network device.

[0601] It should be understood that the first network element can obtain the identifier and / or address information of the second network element in advance to facilitate the first network element sending the first request to the second network element. For example, the address information of the second network element may include the IP address of the second network element, the FQDN of the second network element, or a URL. In addition, the identifier or address information of the second network element described above is merely an example, and the identifier or address information of the second network element may also be other identifiers or address information used for communication with the second network element, and this embodiment of the present application does not specifically limit this.

[0602] In addition, there are two implementations for the first network element to obtain the identifier and / or address information of the second network element in ...

Claims

1. An information transmission method, characterized in that: The method comprises: The first network element sends a first request to the second network element, where the first request includes first information, where the first information is used to indicate at least one first sample alignment algorithm supported by the first network element; The first network element receives a first response from the second network element, where the first response is used to indicate a second sample alignment algorithm among the at least one first sample alignment algorithm, where the second sample alignment algorithm is used to determine common samples between a data set of the first network element for longitudinal federated learning model training and a data set of the second network element for longitudinal federated learning model training; or, the first response is used to indicate that sample alignment is not supported.

2. The method according to claim 1, characterized in that The method further comprises: The first network element receives information about the second network element from a third network element, where the information about the second network element includes an identifier of the second network element and / or address information of the second network element.

3. The method according to claim 2, characterized in that The method further comprises: The first network element sends a second request to the third network element, where the second request includes second information, and the second information is used to indicate a sample alignment requirement; The first network element receives the information of the second network element from the third network element, including: the first network element receives a second response from the third network element, and the second response includes the information of the second network element.

4. The method according to claim 3, characterized in that The sample alignment requirement includes at least one of the following: a third sample alignment type, a third sample alignment algorithm, or a third time period for sample alignment.

5. The method according to any one of claims 2 to 4, characterized in that The information of the second network element further includes at least one of the following: a fourth sample alignment type, a fourth sample alignment algorithm, or a fourth time period supporting sample alignment.

6. The method according to any one of claims 2 to 5, characterized in that The method further comprises: The first network element sends a registration request to the third network element, where the registration request includes third information, and the third information is used to indicate at least one of the following: a sample alignment type supported by the first network element, a supported sample alignment algorithm, or a time period for supporting sample alignment.

7. The method according to any one of claims 2 to 6, characterized in that The at least one first sample alignment algorithm is an algorithm commonly supported by the first network element and the second network element.

8. The method according to any one of claims 1 to 7, characterized in that The first information is further used to indicate at least one of the following: the number of samples supported by the first network element for sample alignment, the first sample alignment type supported by the first network element, or at least one first time period in which the first network element supports sample alignment.

9. The method according to claim 8, characterized in that The at least one first sample alignment algorithm is associated with the at least one first time period.

10. The method according to any one of claims 1 to 9, characterized in that The first response is further used to indicate the number of samples supported by the second network element for sample alignment, and / or the second time period supported by the second network element for sample alignment; Alternatively, the first response is further used to indicate the reason why the second network element does not support sample alignment, and / or the maximum number of samples that the second network element expects the first network element to use.

11. The method according to any one of claims 1 to 10, characterized in that The method further comprises: The first network element sends a first message to the second network element, where the first message is used to indicate a target sample alignment algorithm, where the target sample alignment algorithm is a sample alignment algorithm determined from the at least one first sample alignment algorithm according to the second sample alignment algorithm.

12. The method according to claim 11, characterized in that The method further comprises: The first network element sends the first request to a fourth network element; The first network element receives a third response from the fourth network element, where the third response is used to indicate a fifth sample alignment algorithm among the at least one first sample alignment algorithm, and the fifth sample alignment algorithm is used to determine common samples between a dataset for longitudinal federated learning model training of the first network element and a dataset for longitudinal federated learning model training of the fourth network element; The first network element sends a second message to the fourth network element, and the second message is used to indicate the target sample alignment algorithm; wherein, there are multiple fifth sample alignment algorithms and / or second sample alignment algorithms, and the target sample alignment algorithm is a sample alignment algorithm determined from the at least one first sample alignment algorithm based on the second sample alignment algorithm, including: the target sample alignment algorithm is a sample alignment algorithm determined from the at least one first sample alignment algorithm based on the second sample alignment algorithm and the fifth sample alignment algorithm.

13. The method according to claim 11 or 12, characterized in that The first message is further used to indicate a target time period for sample alignment.

14. The method according to any one of claims 1 to 13, characterized in that The method further comprises: In a case where the first network element determines not to perform sample alignment, the first network element sends a third message to the second network element, where the third message is used to indicate that sample alignment is not to be performed.

15. The method according to claim 14, characterized in that The third message is further used to indicate the reason why sample alignment is not performed.

16. An information transmission method, characterized in that: The method comprises: The first network element sends a first request to the second network element, where the first request includes first indication information, where the first indication information is used to indicate providing a sample identifier for vertical federated learning model training; The first network element receives a first response from the second network element, where the first response includes a first sample identifier.

17. The method according to claim 16, characterized in that The first indication information is used to indicate the provision of sample identifiers for longitudinal federated learning model training, including: the first indication information is used to indicate the provision of sample identifiers corresponding to the data set used for the longitudinal federated learning model training, and the first sample identifier is the sample identifier corresponding to the data set used for the longitudinal federated learning model training from the second network element.

18. The method according to claim 16 or 17, characterized in that The second network element is a candidate network element for participating in the vertical federated learning model training; the method further includes: The first network element receives a second sample identifier corresponding to a data set used for training the vertical federated learning model from a third network element, where the third network element is another network element other than the second network element among multiple candidate network elements participating in the training of the vertical federated learning model; The first network element sends a third message to the second network element based on the first sample identifier and the second sample identifier, where the third message includes second indication information, where the second indication information is used to instruct the second network element to participate in the vertical federated learning model training, and / or a third sample identifier, where the third sample identifier is a common sample identifier of at least two network elements participating in the vertical federated learning model training, where the at least two network elements are at least two network elements determined from the multiple candidate network elements based on the first sample identifier and the second sample identifier, and the at least two network elements include the third network element; Alternatively, the second indication information is used to indicate that the second network element does not participate in the vertical federated learning model training, and / or the reason why the second network element does not participate in the vertical federated learning model training.

19. The method according to claim 16, wherein The second network element is a network element for registering sample identifiers corresponding to data sets of network elements participating in longitudinal federated learning model training; The first indication information is used to indicate the provision of sample identification for longitudinal federated learning model training, including: the first indication information is used to indicate the provision of a common sample identification, the common sample identification is a common sample identification between the data sets of at least two network elements in multiple network elements participating in the same longitudinal federated learning model training, and the first sample identification is used to indicate the common sample between the data sets of at least two network elements in multiple network elements participating in the first longitudinal federated learning model training.

20. The method according to claim 19, characterized in that The first network element is a network element participating in the vertical federated learning model training; the method further includes: The first network element sends a registration request to the second network element, where the registration request includes fourth indication information, and the fourth indication information is used to indicate a sample identifier corresponding to a data set used by the first network element for training a longitudinal federated learning model.

21. The method according to claim 19 or 20, characterized in that The first request further includes fifth indication information and / or sixth indication information, the fifth indication information is used to indicate the quantity range of sample identifiers included in the common sample identifier, and the sixth indication information is further used to instruct the second network element to determine the at least two network elements.

22. The method according to any one of claims 19 to 21, characterized in that The first response further includes identifiers and / or address information of the at least two network elements, and / or the number of sample identifiers included in the first sample identifier.

23. A communication device, characterized in that: The communication device includes a module or unit for executing the method according to any one of claims 1 to 15, or includes a module or unit for executing the method according to any one of claims 16 to 22.

24. A communication device, characterized in that: The communication device includes a processor, and the processor is configured to enable the communication device to perform the method according to any one of claims 1 to 15 through logic circuits and / or execution instructions, or enable the communication device to perform the method according to any one of claims 16 to 22.

25. The communication device according to claim 24, characterized in that The communication device further includes a memory configured to store the instruction.

26. The communication device according to claim 24 or 25, characterized in that The communication device further comprises a communication interface, which is used to input and / or output signaling and / or data.

27. A computer-readable storage medium, characterized in that The computer-readable storage medium comprises instructions, and when the instructions are executed by a processor, the method according to any one of claims 1 to 15 is implemented, or the method according to any one of claims 16 to 22 is implemented.

28. A computer program product, characterized in that The computer program product comprises instructions, which, when executed on a computer, cause the computer to perform the method according to any one of claims 1 to 15, or cause the computer to perform the method according to any one of claims 16 to 22.

Images