Sample alignment method and apparatus in vertical federated learning, and readable storage medium

Abstract

The present application relates to the technical field of communications, and in particular to a sample alignment method and apparatus in vertical federated learning (VFL), and a readable storage medium. The method comprises: a first network element sends a first request to a second network element for requesting a sample, wherein the first request comprises the identifier type of a sample used in a VFL task; the first network element receives a first response from the second network element, wherein the first response comprises the identifier of a sample supported by the second network element, and the identifier of the sample supported by the second network element matches the identifier type of the sample used in the VFL task; and the first network element determines, on the basis of the identifier of the sample supported by the second network element, the identifier of a sample supported by a group of VFL tasks jointly, wherein the group of VFL tasks comprises the second network element, and optionally comprises the first network element. By means of the present application, the possibility of sample identifier mismatch in a sample alignment process of a VFL task can be reduced, achieving sample alignment in the VFL task.

Description

Sample alignment method and device in vertical federated learning and readable storage medium

[0001] The present application claims priority to the Chinese patent application No. 202410727045.2, filed on June 5, 2024, entitled "Sample alignment method and device in vertical federated learning and readable storage medium", the whole content of which is incorporated herein by reference. TECHNICAL FIELD

[0002] The present application relates to the field of communication technology, in particular to a sample alignment method and device in vertical federated learning and readable storage medium. BACKGROUND

[0003] Participating entities (for example, network elements in a vertical federated alliance) of a vertical federated learning (VFL) task need to align the sample conditions of participating in the vertical federated learning task by different entities, that is, the same range of data is trained and / or inferred by different entities. At present, the 3rd generation partnership project (3GPP) supports training and / or inference through a vertical federated learning task to realize mobile network related analysis services.

[0004] At present, how the participating entities of the vertical federated learning align the samples in the vertical federated learning task is a problem that technicians in the field are studying. SUMMARY

[0005] The embodiments of the present application provide a sample alignment method and device in vertical federated learning and readable storage medium, which can reduce the possibility of sample identification mismatch in the sample alignment process of the vertical federated learning task, and realize sample alignment in the vertical federated learning task.

[0006] The present application will be described from different aspects below. It should be understood that the implementation and advantages of the different aspects below can be mutually referred to.

[0007] In a first aspect, a sample alignment method in vertical federated learning is provided. The method can be performed by a communication device, which can be a first network element or a chip in the first network element. For ease of description, the communication device is taken as the first network element in the description. The method includes: sending, by the first network element, a first request to a second network element, the first request being used to request samples, and the first request including an identification type of samples used in a VFL task; receiving, by the first network element, a first response from the second network element, the first response including an identification of samples supported by the second network element, and the identification of samples supported by the second network element matching the identification type of samples used in the VFL task; and determining, by the first network element, an identification of samples commonly supported by a VFL task group based on the identification of samples supported by the second network element.

[0008] For example, the identification of samples supported by the second network element can be encrypted (in ciphertext form) or unencrypted (in plaintext form). For details, see the description of the embodiments below, which are not described here.

[0009] For example, the first network element can be a network data analytics function (NWDAF), the second network element can be an application function (AF), and the VFL task group includes the first network element and the second network element. Alternatively, the first network element can be an AF, the second network element can be an NWDAF, and the VFL task group includes the first network element and the second network element. Alternatively, the first network element can be a network exposure function (NEF), the second network element can be an AF or an NWDAF, and the VFL task group includes the second network element.

[0010] For example, the identification type of samples used in the VFL task can be a subscription permanent identifier (SUPI), a generic public subscription identifier (GPSI), or an IP address.

[0011] For example, the identification of samples matching the identification type of samples used in the VFL task can mean that the type of the identification of samples is the same as the identification type of samples used in the VFL task, such as both being SUPI, or both being GPSI, or both being an IP address.

[0012] In this application, the identification type of the sample used in the VFL task can be understood as: the identification type of the sample used in the sample alignment process of the VFL task. In the subsequent training process of the VFL task, the identification type of the sample can be updated. In other words, the identification type of the sample used in the sample alignment process of the VFL task can be different from the identification type of the sample used in the subsequent training process, of course, it can also be the same, and the embodiments of the present application do not make any limitation.

[0013] For example, the identification type of the sample used in the VFL task can include one or more of the intersection between the identification of the sample supported by the first network element (NWDAF or AF) and the identification of the sample supported by the second network element (AF or NWDAF). It can be understood that there can be multiple second network elements, and if there are multiple second network elements, the identification type of the sample used in the VFL task can include one or more of the intersection between the identification of the sample supported by the first network element (NWDAF or AF) and the identification of the sample supported by all second network elements (AF or NWDAF).

[0014] The present application carries the identification type of the sample used in the VFL task in the first request, so that the second network element returns the sample identification matched with the identification type of the sample used in the VFL task, which can reduce the possibility of sample identification mismatch in the sample alignment process of the VFL task, and realize the sample alignment in the VFL task.

[0015] In this application, the sample supported by a network element can be understood as the sample that can be used by the network element for machine learning, artificial intelligence tasks (training and / or inference), for example, vertical federated training tasks.

[0016] In combination with the first aspect, in a possible implementation, before the first network element sends the first request to the second network element, the method further includes: the first network element and the second network element respectively register their capabilities to a network repository function (NRF) network element.

[0017] For example, the first network element can send a network element registration request to the NRF network element to register its own capabilities. The network element registration request can include the identification type of the sample supported by the first network element. The NRF network element stores the information in the network element registration request, and returns a network element registration response to the first network element to confirm that the registration of the first network element is accepted. Similarly, the second network element can send a network element registration request to the NRF network element to register its own capabilities. The network element registration request can include the identification type of the sample supported by the second network element. The NRF network element stores the information in the network element registration request, and returns a network element registration response to the second network element to confirm that the registration of the second network element is accepted.

[0018] With reference to the first aspect, in a possible implementation manner, before the first network element sends the first request to the second network element, the method further includes: determining, by the first network element, the identification type of the sample used in the VFL task.

[0019] For example, the first network element sends a network element discovery request to a network repository function (NRF) network element, for discovering or screening the second network element. The network element discovery request can include, but is not limited to, an expected network function (NF) service name, an NF type of an expected NF instance, or an NF type of an NF consumer, etc. After receiving the network element discovery request, the NRF network element can select one or more second network elements based on the information carried in the network element discovery request, and can return a network element discovery response to the first network element. The network element discovery response includes information of the second network element, such as the identification type of the sample supported by the second network element, and optionally, an identifier of the second network element. After receiving the network element discovery response, the first network element can determine the identification type of the sample used in the VFL task based on the identification type of the sample supported by the first network element and / or the identification type of the sample supported by the second network element. The identification type of the sample used in the VFL task can include one or more of the intersection between the identification type of the sample supported by the first network element and the identification type of the sample supported by the second network element.

[0020] For example, the first network element sends a network element discovery request to the NRF network element, and the network element discovery request can include the identification type of the sample supported by the first network element. The first network element receives a network element discovery response from the NRF network element, and the network element discovery response can include information of the second network element, such as the intersection between the identification type of the sample supported by the second network element and the identification type of the sample supported by the first network element, and optionally, an identifier of the second network element. After receiving the network element discovery response, the first network element can take the intersection in the network element discovery response as the identification type of the sample used in the VFL task. Further, after receiving the network element discovery request, the NRF network element can select one or more second network elements based on the information carried in the network element discovery request, and the second network elements support the intersection between the identification type of the sample supported by the first network element and the identification type of the sample supported by the first network element, and the intersection is not empty, and the NRF network element can return a network element discovery response to the first network element.

[0021] Exemplarily, the first network element sends a network element discovery request to the NRF network element, and the network element discovery request can further include an identification type of a sample expected to be supported. Here, the identification type of the sample expected to be supported can be part or all of the identification types of the samples supported by the first network element (NWDAF), or can be the identification types of the samples not supported by the first network element (NWDAF), and the embodiments of the present application are not limited. The first network element receives a network element discovery response from the NRF network element, and the network element discovery response can include information of the second network element, for example, an identifier of the second network element. The second network element supports the identification type of the sample expected to be supported in the network element discovery request. The first network element can use the identification type of the sample expected to be supported as the identification type of the sample used in the VFL task. Further, after receiving the network element discovery request, the NRF network element can filter one or more second network elements supporting the identification type of the sample expected to be supported from the locally stored information, and can return the network element discovery response to the first network element.

[0022] In the network element registration process, each network element of the present application registers the identification types of the samples supported by the network element to the NRF network element, and then selects a VFL participant (for example, the second network element) and determines the identification types of the samples supported by the VFL participant through the network element discovery process, so as to determine the identification type of the sample used in the VFL task. The identification type of the sample used in the VFL task can be determined under the existing process, without the need to design a new process, and the complexity is low.

[0023] In combination with the first aspect, in a possible implementation manner, the first request further includes an encrypted identification of a sample supported by the first network element; and the identification of the sample supported by the second network element is an intersection between the encrypted identification of the sample supported by the second network element and the encrypted identification of the sample supported by the first network element.

[0024] Exemplarily, before the first network element sends the first request to the second network element, the method further includes: when the sample identifier a stored by the first network element does not match the identification type of the sample used in the VFL task, the first network element converts the sample identifier a into a sample identifier b, and performs encryption processing on the sample identifier b to obtain the encrypted identification of the sample supported by the first network element. The sample identifier b matches the identification type of the sample used in the VFL task. When the sample identifier a stored by the first network element matches the identification type of the sample used in the VFL task, the first network element performs encryption processing on the sample identifier a to obtain the encrypted identification of the sample supported by the first network element. Here, the encryption processing can be based on encryption information, which can be predefined or preconfigured. For example, the encryption information can include an encryption algorithm type, and / or specific information of the encryption algorithm (such as a public key, encryption, etc.).

[0025] After receiving the first request, the second network element can determine whether the locally stored sample identifier c matches the identifier type of the sample used in the VFL task carried in the first request. If the sample identifier c stored locally by the second network element does not match the identifier type of the sample used in the VFL task, the second network element can convert the sample identifier c to a sample identifier d that matches the identifier type of the sample used in the VFL task. The second network element can determine encryption information (such as carried in the first request, or predefined or preconfigured), and can use the encryption information to encrypt the sample identifier d to obtain an encrypted identifier of a sample supported by the second network element. If the sample identifier c stored locally by the second network element matches the identifier type of the sample used in the VFL task, the second network element can determine encryption information (such as carried in the first request, or predefined or preconfigured), and can use the encryption information to encrypt the sample identifier c to obtain an encrypted identifier of a sample supported by the second network element. The second network element can then determine the intersection between the encrypted identifier of the sample supported by the first network element in the first request and the encrypted identifier of the sample supported by the second network element, and can send a first response to the first network element. The first response can include the intersection between the encrypted identifier of the sample supported by the first network element and the encrypted identifier of the sample supported by the second network element.

[0026] The encryption algorithm in the present application can be homomorphic encryption. Homomorphic encryption refers to performing a specific operation on the ciphertext obtained after homomorphic encryption of the original data, and then performing homomorphic decryption on the calculation result to obtain the plaintext, which is equivalent to the result obtained by directly performing the same operation on the original data (plaintext).

[0027] For example, the first request can further include the encryption information for the second network element to encrypt the identifier of the sample.

[0028] The first network element and the second network element in the present application encrypt the identifiers of the samples supported by each of them, and the identifiers of the samples before encryption match the identifier type of the sample in the VFL task. Therefore, even if the other party cannot or does not obtain the plaintext, the intersection of the encrypted ciphertext can be used to obtain the identifiers of the samples supported by the first network element and the second network element. Not only can the security be improved (by encryption), but also the possibility of sample identifier mismatch in the sample alignment process of the vertical federated learning task can be reduced (this effect is achieved by aligning the identifier type of the sample before encryption), thereby achieving sample alignment in the vertical federated learning task.

[0029] In a possible implementation manner of the first aspect, the first request can further include an identifier of the first network element, and the identifier of the sample supported by the second network element is a plaintext identifier. After receiving the first request, the second network element can determine whether the first request is used to request a plaintext identifier of a sample or a ciphertext identifier of a sample. If the first request is used to request a plaintext identifier of a sample, the second network element can determine, based on a type of the first network element, whether to return the plaintext identifier of the sample. For example, if the first network element (i.e., the requester of the sample) is a NWDAF or a trusted AF (trusted AF), the second network element can agree to return the plaintext identifier of the sample. If the first network element (i.e., the requester of the sample) is an untrusted AF (untrusted AF) or an AF, the second network element can not send the plaintext identifier of the supported sample to the first network element, and in this case, the second network element can refuse the request of the first network element, and of course, can return the encrypted ciphertext identifier, which is not limited by the present application. When the second network element agrees to return the plaintext identifier of the sample, the second network element sends the plaintext identifier of the sample supported by the second network element to the first network element in the first response. Here, the plaintext identifier of the sample supported by the second network element is matched with the type of the identifier of the sample used in the VFL task.

[0030] In a possible implementation manner of the first aspect, the identifier of the sample supported by the second network element is an encrypted identifier of the sample supported by the second network element.

[0031] For example, the first network element determines the identifier of the sample commonly supported by the VFL task group based on the identifier of the sample supported by the second network element, including that the first network element determines the identifier of the sample commonly supported by the VFL task group based on the encrypted identifier of the sample supported by the first network element and the encrypted identifier of the sample supported by the second network element. The identifier of the sample commonly supported by the VFL task group can be a plaintext identifier corresponding to an intersection of the encrypted identifier of the sample supported by the first network element and the encrypted identifier of the sample supported by the second network element. It can be understood that the identifier of the sample commonly supported by the VFL task group can be matched with the type of the identifier of the sample used in the VFL task.

[0032] The identifier of the sample returned by the second network element is matched with the type of the identifier of the sample used in the VFL task, and the identifier of the sample is a ciphertext. Even if the first network element cannot or does not obtain the plaintext, the first network element can use the encrypted ciphertext of the sample supported by the first network element and matched with the type of the identifier of the sample used in the VFL task to take an intersection with the identifier of the sample (ciphertext) returned by the second network element, to obtain the identifier of the sample commonly supported by the first network element and the second network element. This can not only improve security (achieved by encryption), but also reduce the possibility of sample identifier mismatch in the sample alignment process of the vertical federated learning task (achieved by aligning the type of the identifier of the sample before encryption), thereby achieving sample alignment in the vertical federated learning task.

[0033] In a possible implementation of the first aspect, in the method, the first network element is an NEF, and the second network element is an AF or an NWDAF. Before the first network element sends the first request to the second network element, the method further includes: the first network element receives a second request from a third network element, the second request including one or more of the following: a sample identification type used in a vertical federated learning (VFL) task, information used to determine the second network element (such as an identifier or a name of the second network element), or role information corresponding to the second network element (for example, a VFL active participant or a VFL passive participant). The sample identification type used in the VFL task can be preconfigured, predefined, or specified in a standard protocol. The sample identification type used in the VFL task can also be determined through a network element discovery process, or determined by the third network element from sample identification types (such as a SUPI, a GPSI, or an IP address) supported by the third network element. The information used to determine the second network element can be obtained through a network element discovery process. For details, refer to the description of the embodiments below, which are not described here due to space limitations.

[0034] For example, the first request can further include one or more of the following: role information corresponding to the second network element (for example, a VFL active participant or a VFL passive participant), an analysis identifier, or encryption information. The encryption information can include an encryption algorithm type, and / or specific information of the encryption algorithm (such as a public key, encryption, and the like). The encryption algorithm can be a homomorphic encryption (Homomorphic Encryption).

[0035] For example, the third network element can be an NWDAF.

[0036] For example, the second request can further include an identification of a sample supported by the third network element. The first network element determines an identification of a sample commonly supported by a VFL task group based on the identification of the sample supported by the second network element, including: the first network element determines the identification of the sample commonly supported by the VFL task group based on the identification of the sample supported by the third network element and the identification of the sample supported by the second network element, the VFL task group including the second network element and the third network element.

[0037] It can be understood that if the identification of the sample supported by the second network element in the first response is in plaintext, and the identification of the sample supported by the third network element is also in plaintext, then the identification of the sample commonly supported by the VFL task group can be an intersection between the identification of the sample supported by the third network element and the identification of the sample supported by the second network element. It can be understood that the identification of the sample commonly supported by the VFL task group can match the sample identification type used in the VFL task included in the first request.

[0038] In a second aspect, the present application provides a sample alignment method in vertical federated learning, which can be executed by a communication device, which can be a second network element or a chip in the second network element. For ease of description, the communication device is taken as the second network element for example. The method comprises: the second network element receives a first request from a first network element, the first request comprising an identification type of a sample used in a VFL task; and the second network element sends a first response to the first network element, the first response comprising an identification of a sample supported by the second network element, which matches the identification type of the sample used in the VFL task.

[0039] For example, the identification of the sample supported by the second network element can be encrypted (in ciphertext form) or unencrypted (in plaintext form). For details, see the description of the embodiments below, which are not described here.

[0040] For example, the first network element can be a NWDAF, the second network element can be an AF, and the VFL task group comprises the first network element and the second network element. Alternatively, the first network element can be an AF, the second network element can be a NWDAF, and the VFL task group comprises the first network element and the second network element. Alternatively, the first network element can be a NEF, the second network element can be an AF or a NWDAF, and the VFL task group comprises the second network element.

[0041] For example, the identification type of the sample used in the VFL task can be SUPI, GPSI, or IP address.

[0042] For example, the identification of the sample matches the identification type of the sample used in the VFL task, which means that the type of the identification of the sample is the same as the identification type of the sample used in the VFL task, such as both are SUPI, or both are GPSI, or both are IP address.

[0043] For example, after receiving the first response, the second network element can determine the identification of the sample supported by the VFL task group based on the identification of the sample supported by the second network element in the first response.

[0044] The first request of the present application carries the identification type of the sample used in the VFL task. After receiving the first request, the second network element can return the sample identification matching the identification type of the sample used in the VFL task, thereby reducing the possibility of sample identification mismatch in the sample alignment process of the VFL task and realizing the sample alignment in the VFL task.

[0045] In a possible implementation of the second aspect, the second network element can send a network element registration request to the NRF network element, for registering its capability. The network element registration request can include the identification type of the samples supported by the second network element. The NRF network element stores the information in the network element registration request, and returns a network element registration response to the second network element, for confirming that the registration of the second network element is accepted.

[0046] In a possible implementation of the second aspect, the first request further includes the encrypted identification of the samples supported by the first network element; and the identification of the samples supported by the second network element is the intersection between the encrypted identification of the samples supported by the second network element and the encrypted identification of the samples supported by the first network element.

[0047] For example, before the second network element sends the first response to the first network element, the method further includes: determining, by the second network element, the encrypted identification of the samples supported by the second network element that matches the identification type of the samples used in the VFL task, according to the identification type of the samples used in the VFL task.

[0048] For example, after the second network element receives the first request from the first network element, the second network element can determine whether the sample identification c stored locally matches the identification type of the samples used in the VFL task carried in the first request. If the sample identification c stored locally by the second network element does not match the identification type of the samples used in the VFL task, the second network element can convert the sample identification c to sample identification d that matches the identification type of the samples used in the VFL task. The second network element can perform encryption processing on the sample identification d to obtain the encrypted identification of the samples supported by the second network element. If the sample identification c stored locally by the second network element matches the identification type of the samples used in the VFL task, the second network element can perform encryption processing on the sample identification c to obtain the encrypted identification of the samples supported by the second network element. The second network element can then determine the intersection between the encrypted identification of the samples supported by the first network element and the encrypted identification of the samples supported by the second network element in the first request, and send the first response to the first network element. The first response can include the intersection between the encrypted identification of the samples supported by the first network element and the encrypted identification of the samples supported by the second network element.

[0049] The encryption processing can be based on encryption information. The encryption information can include an encryption algorithm type, and / or specific information of the encryption algorithm (such as a public key, encryption, etc.). The encryption algorithm can be a homomorphic encryption.

[0050] Exemplarily, the encryption information can be pre-configured or pre-defined or standard specified, etc., or can be determined in advance by the first network element and the second network element. Alternatively, the encryption information can also be carried in the first request. In other words, the first request can also include the encryption information.

[0051] The first network element and the second network element of the present application encrypt the identification of the respective supported samples, and the identification of the samples before encryption is matched with the identification type of the samples in the VFL task. Therefore, even if the other party cannot or does not obtain the plaintext, the identification of the samples supported by the first network element and the second network element can be obtained by using the ciphertext after encryption. Not only the security can be improved (achieved by encryption), but also the possibility of sample identification mismatch in the sample alignment process of the vertical federated learning task can be reduced (this effect is achieved by aligning the identification type of the samples before encryption), thereby realizing the sample alignment in the vertical federated learning task.

[0052] In combination with the second aspect, in a possible implementation manner, the first request can also include the identification of the first network element, and the identification of the samples supported by the second network element can be the plaintext identification. After receiving the first request, the second network element can determine whether to return the plaintext identification of the samples based on the type of the first network element. For example, if the first network element (i.e., the requester of the samples) is a NWDAF or a trusted AF (trusted AF), the second network element can agree to return the plaintext identification of the samples. If the first network element (i.e., the requester of the samples) is an untrusted AF (untrusted AF) or an AF, the second network element can not send the plaintext identification of the supported samples to the first network element, and in this case, the second network element can refuse the request of the first network element, and of course can return the ciphertext identification after encryption, which is not limited by the present application. When the second network element agrees to return the plaintext identification of the samples, the second network element carries the plaintext identification of the samples supported by the second network element in the first response and sends the first network element. Here, the plaintext identification of the samples supported by the second network element is matched with the identification type of the samples used in the VFL task.

[0053] Exemplarily, after receiving the first request, the second network element can first determine whether the first request is used to request the plaintext identification of the samples or the ciphertext. If the first request is used to request the plaintext identification of the samples, the second network element can then determine whether to return the plaintext identification of the samples based on the type of the first network element.

[0054] In combination with the second aspect, in a possible implementation manner, the identification of the samples supported by the second network element is the encrypted identification of the samples supported by the second network element.

[0055] With reference to the second aspect, in a possible implementation manner, the first network element is an NEF, and the second network element is an AF or an NWDAF. The identifier of the sample supported by the second network element is a sample identifier c (in plaintext) stored by the second network element, and the sample identifier c matches an identifier type of the sample used in the VFL task.

[0056] With reference to the third aspect, the present application provides a sample alignment method in vertical federated learning. The method can be executed by a communication apparatus, which can be a third network element or a chip in the third network element. For ease of description, the communication apparatus is taken as the third network element in the following description. The method comprises: determining, by the third network element, an identifier type of a sample used in a VFL task; and sending, by the third network element, a second request to a first network element, the second request comprising one or more of the following: the identifier type of the sample used in the VFL task, information used to determine the second network element, or role information (for example, a VFL active participant or a VFL passive participant) corresponding to the second network element. The information used to determine the second network element can comprise an identifier or a name of the second network element, and can also be other information. Any information that can determine the second network element is within the protection scope of the present application. The information used to determine the second network element can be obtained through a network element discovery process.

[0057] For example, the third network element can be an NWDAF, the first network element can be an NEF, and the second network element can be an AF. Alternatively, the third network element can be an AF, the first network element can be an NEF, and the second network element can be an NWDAF.

[0058] For example, the identifier type of the sample used in the VFL task can be a SUPI, a GPSI, or an IP address.

[0059] For example, the second request can further comprise encryption information used by the second network element for encryption. The encryption information can comprise an encryption algorithm type and / or specific information of the encryption algorithm (for example, a public key, encryption, and the like). The encryption algorithm can be a homomorphic encryption.

[0060] With reference to the third aspect, in a possible implementation manner, the second request further comprises an identifier of a sample supported by the third network element. The identifier of the sample supported by the third network element can be encrypted or unencrypted. If the identifier of the sample supported by the third network element is encrypted, the third network element uses the encryption information for encryption.

[0061] For example, the identifier of the sample supported by the third network element can match the identifier type of the sample used in the VFL task.

[0062] In a possible implementation manner of the third aspect, before the third network element sends the second request to the first network element, the method further includes: the third network element and the second network element respectively register their capabilities to an NRF network element.

[0063] For example, the third network element can send a network element registration request to the NRF network element, for registering its own capability. The network element registration request can include the identification types of samples supported by the third network element. The NRF network element stores the information in the network element registration request, and returns a network element registration response to the third network element, for confirming that the registration of the third network element is accepted. Similarly, the second network element can send a network element registration request to the NRF network element, for registering its own capability. The network element registration request can include the identification types of samples supported by the second network element. The NRF network element stores the information in the network element registration request, and returns a network element registration response to the second network element, for confirming that the registration of the second network element is accepted.

[0064] In a possible implementation manner of the third aspect, the third network element determines the identification types of samples used in the vertical federated learning task, including: the third network element sends a network element discovery request to the NRF network element, for discovering or screening the second network element. The network element discovery request can include but is not limited to: an expected NF service name, an expected NF type of NF instance, or an NF type of NF consumer, etc. After receiving the network element discovery request, the NRF network element can select one or more second network elements based on the information carried in the network element discovery request, and can return a network element discovery response to the third network element. The network element discovery response includes the information of the second network element, such as the identification types of samples supported by the second network element, and optionally includes the identification of the second network element. After receiving the network element discovery response, the third network element can determine the identification types of samples used in the VFL task based on the identification types of samples supported by the third network element, and the identification types of samples supported by the second network element. The identification types of samples used in the VFL task can include one or more in the intersection between the identification types of samples supported by the first network element and the identification types of samples supported by the second network element.

[0065] Exemplarily, the network element discovery request can further include an identification type of a sample supported by the third network element. After receiving the network element discovery request, the NRF network element can select one or more second network elements based on the information carried in the network element discovery request, the second network elements support an intersection between the identification type of the sample supported by the third network element and the identification type of the sample, and the intersection is not empty, and can return a network element discovery response to the third network element. The network element discovery response can include information of the second network element, for example, the intersection between the identification type of the sample supported by the second network element and the identification type of the sample supported by the third network element, and optionally, the identification of the second network element. After receiving the network element discovery response, the third network element can take the intersection in the network element discovery response as the identification type of the sample used in the VFL task.

[0066] Exemplarily, the network element discovery request can further include an identification type of a sample expected to be supported. After receiving the network element discovery request, the NRF network element can filter one or more second network elements supporting the identification type of the sample expected to be supported from the locally stored information, and can return a network element discovery response to the third network element. The network element discovery response can include information of the second network element, for example, the identification of the second network element. The third network element can take the identification type of the sample expected to be supported as the identification type of the sample used in the VFL task.

[0067] In a fourth aspect, a communication apparatus is provided. The communication apparatus can be the first network element, or the second network element, or the third network element, or a chip therein. The communication apparatus includes a unit and / or module for performing the method provided in any of the first aspect to the third aspect, or any possible implementation manner of any of the first aspect to the third aspect, for example, a transceiver module and / or a processing module. The transceiver module is used to transceive various information or signaling, and thus can also achieve the beneficial effects (or advantages) possessed by the method provided in any of the first aspect to the third aspect.

[0068] In a possible implementation manner, the communication apparatus can be the first network element or a chip in the first network element. The transceiver module is configured to send a first request to the second network element, the first request including an identification type of a sample used in a vertical federated learning task; the transceiver module is further configured to receive a first response from the second network element, the first response including an identification of a sample supported by the second network element, the identification of the sample supported by the second network element matching the identification type of the sample used in the vertical federated learning task; and the processing module is configured to determine an identification of a sample commonly supported by a vertical federated learning task group based on the identification of the sample supported by the second network element, the vertical federated learning task group including the second network element.

[0069] In a possible implementation, the communication apparatus can be a second network element or a chip in the second network element. The transceiver is configured to receive a first request from a first network element, the first request comprising an identification type of samples used in a federated learning task in a vertical direction; and the transceiver is further configured to send a first response to the first network element, the first response comprising an identification of samples supported by the second network element, the identification of samples supported by the second network element matching the identification type of samples used in the federated learning task in the vertical direction.

[0070] For example, the processing module is configured to generate various information sent by the transceiver, such as the first request; and the processing module is further configured to control the transceiver to send or receive various information.

[0071] In a possible implementation, the communication apparatus can be a third network element or a chip in the third network element. The processing module is configured to determine an identification type of samples used in a federated learning task in a vertical direction; and the transceiver is configured to send a second request to a first network element, the second request comprising information for determining the second network element and the identification type of samples used in the federated learning task in the vertical direction.

[0072] In a fifth aspect, a communication apparatus is provided, which comprises a processor configured to execute the method in any of the first aspect to the third aspect, or any possible implementation of any of the first aspect to the third aspect. Alternatively, the processor is configured to execute a program stored in a memory, and when the program is executed, the method in any of the first aspect to the third aspect, or any possible implementation of any of the first aspect to the third aspect is executed.

[0073] With reference to the fifth aspect, in a possible implementation, the memory is located outside the communication apparatus.

[0074] With reference to the fifth aspect, in a possible implementation, the memory is located inside the communication apparatus.

[0075] In this application, the processor and the memory can also be integrated into one device, that is, the processor and the memory can also be integrated together.

[0076] With reference to the fifth aspect, in a possible implementation, the communication apparatus further comprises a transceiver configured to send or receive various information, for example, receiving the first request, sending the first response, and the like.

[0077] In a sixth aspect, the present application provides a communication device, which can include a processor and an interface circuit, which are connected. Wherein the interface circuit is used to interact (or transceive or input and output) information or data, and the processor is used to run program instructions, so that the communication device executes the method described in any one of the first aspect to the third aspect, or any possible implementation manner of any one of the aspects. Wherein the interface circuit can be a communication interface, or a transceiver. The transceiver can be a radio frequency module in the communication device, or a combination of a radio frequency module and an antenna, or an input and output interface of a chip or circuit.

[0078] In a possible implementation manner, the interface circuit is configured to output the first request; the interface circuit is configured to input the first response; and the processor is configured to determine the identification of the samples commonly supported by the longitudinal federated learning task group based on the identification of the samples supported by the second network element.

[0079] In another possible implementation manner, the interface circuit is configured to input the first request; and the interface circuit is configured to output the first response.

[0080] In still another possible implementation manner, the processor is configured to determine the identification type of the samples used in the longitudinal federated learning task; and the interface circuit is configured to output the second request.

[0081] Wherein, the specific description about the first request, the first response, the identification of the samples, the identification of the samples commonly supported by the longitudinal federated learning task group, the identification type of the samples used in the longitudinal federated learning task, and the second request can refer to the method embodiments below, which will not be repeated here.

[0082] In a seventh aspect, the present application provides a readable storage medium, which stores program instructions, when the program instructions are run on a communication device, the communication device executes the method described in any one of the first aspect to the third aspect, or any possible implementation manner of any one of the aspects.

[0083] In an eighth aspect, the present application provides a program product containing instructions, when the instructions are run, the model authorization method described in any one of the possible implementation manners of the first aspect to the third aspect is executed.

[0084] In a ninth aspect, the present application provides a communication apparatus, which can be in the form of a chip or a device. The apparatus includes a processor. The processor is configured to read and execute a program stored in a memory, so as to execute the model authorization method in any one of the first aspect to the third aspect, or any possible implementation of any one of the aspects. Optionally, the apparatus further includes the memory, which is connected to the processor by an electrical circuit. Further optionally, the apparatus further includes a communication interface, which is connected to the processor. The communication interface is configured to receive information and / or signaling to be processed. The processor acquires the information and / or signaling from the communication interface, processes the information and / or signaling, and outputs the processing result through the communication interface. The communication interface can be an input / output interface.

[0085] Optionally, the processor and the memory can be physically independent units, or the memory can be integrated with the processor.

[0086] In a tenth aspect, the present application provides a communication system, which includes one or more network elements of the first network element, the second network element, or the third network element. The first network element can be configured to execute the method described in the first aspect or any possible implementation of the first aspect. The second network element can be configured to execute the method described in the second aspect or any possible implementation of the second aspect. The third network element can be configured to execute the method described in the third aspect or any possible implementation of the third aspect.

[0087] The technical effects achieved by the above aspects can be referred to each other or the beneficial effects in the method embodiments shown below, which will not be described here. BRIEF DESCRIPTION OF DRAWINGS

[0088] FIG. 1 is a schematic diagram of a 5G system architecture according to an embodiment of the present application;

[0089] FIG. 2 is a schematic diagram of a process of NF service registration according to an embodiment of the present application;

[0090] FIG. 3 is a schematic diagram of a process of NF service update according to an embodiment of the present application;

[0091] FIG. 4 is a schematic diagram of a process of NF discovery / NF service discovery according to an embodiment of the present application;

[0092] FIG. 5 is a schematic diagram of a sample alignment process according to an embodiment of the present application;

[0093] FIG. 6 is a schematic diagram of another sample alignment process according to an embodiment of the present application;

[0094] FIG. 7 is a flow diagram of a sample alignment method in federated learning according to an embodiment of the present application;

[0095] FIG. 8 is another flow diagram of a sample alignment method in federated learning according to an embodiment of the present application;

[0096] FIG. 9 is yet another flow diagram of a sample alignment method in federated learning according to an embodiment of the present application;

[0097] FIG. 10 is a structural diagram of a communication device according to an embodiment of the present application;

[0098] FIG. 11 is another structural diagram of a communication device according to an embodiment of the present application;

[0099] FIG. 12 is yet another structural diagram of a communication device according to an embodiment of the present application. DETAILED DESCRIPTION

[0100] In the description of the present application, unless otherwise specified, " / " means "or", for example, A / B can mean A or B. "And / or" herein is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can mean: A exists alone, A and B exist together, and B exists alone. In addition, "at least one" means one or more, and "multiple" means two or more. "At least one of the following" or "one or more of the following" or the like means any combination of these items, including any combination of single item or multiple items. For example, at least one of a, b, or c can mean: a, b, c; a and b; a and c; b and c; or a and b and c. Where a, b, and c can be single or multiple.

[0101] In the description of the present application, "first", "second", and the like are only used to distinguish different objects, and do not limit the quantity and execution order, and "first", "second", and the like do not necessarily distinguish. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device, etc. that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products, or devices.

[0102] In this application, the word "exemplary" is used to mean "serving as an example, instance, or illustration." Any embodiment or design described in this application as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments or designs. Rather, use of the word exemplary is intended to present concepts in a concrete manner.

[0103] It should be understood that in this application, "when", "if" and "as" all refer to the device will make corresponding processing under certain objective conditions, not limited time, and also do not require the device to have a judgment action when it is implemented, nor does it mean that there are other limitations.

[0104] In this application, the element expressed by the singular is intended to represent "one or more", not "one and only one", unless otherwise specified.

[0105] In addition, the terms "system" and "network" are often used interchangeably in this document.

[0106] It can be understood that in the embodiments of the present application, "A corresponds to B", "A and B correspond / associate" and the like similar expressions, all represent that A and B exist corresponding relationship, and B can be determined according to A. However, it should also be understood that the determination of B according to A does not mean that B is determined only according to A, but B can also be determined according to A and / or other information.

[0107] The network element and system architecture related to the present application are briefly described below.

[0108] The technical solutions provided by the present application can be applied to various communication systems deployed with virtual network functions. For example: the fifth generation (5th generation, 5G) communication system or new radio (new radio, NR), equipped with network function virtualization infrastructure (network functions virtualization infrastructure, NFVI) or other virtual network function long term evolution (long term evolution, LTE) network, MulteFire network (create a new wireless network by running LTE technology independently on unlicensed spectrum (such as global 5GHz unlicensed spectrum)), or home base station network, wireless fidelity (wireless fidelity, Wi-Fi) access mobile network, wideband code division multiple access (wideband code division multiple access, WCDMA) network, fixed mobile convergence network (fixed access network access mobile network), and other future communication systems, such as the sixth generation mobile communication system.

[0109] For example, the technical solutions provided in the present application can be applied to the 5G system architecture defined in the 3rd generation partnership project technical specifications (3GPP TS) 23.288.

[0110] Referring to FIG. 1, FIG. 1 is a schematic diagram of a 5G system architecture provided by an embodiment of the present application. As shown in FIG. 1, the 5G system architecture 100 includes but is not limited to an access network, a core network (CN), a data network (DN) 140, and an application function (AF) 141. The access network can be used to implement wireless access related functions, which can include a radio access network (RAN) 120 and a user equipment (UE) 110. The core network can include various network functions (NFs) or network elements, for example, including all or part of the following logical functions: a user plane function (UPF) 130, a network exposure function (NEF) 131, a network repository function (NRF) 132, a policy control function (PCF) 133, a unified data management (UDM) function 134, a unified data repository (UDR) function 135, a network data analytics function (NWDAF) 136, an authentication server function (AUSF) 137, an access and mobility management function (AMF) 138, or a session management function (SMF) 139, etc.

[0111] In a possible implementation, the UE can access a data network by establishing a session between the UE, a RAN, a UPF, and a data network (DN), i.e., a protocol data unit (PDU) session.

[0112] The UE can be a terminal device, such as a mobile phone, an internet of things terminal device, a smart terminal, a vehicle terminal, a vehicle device, a wearable device, a multimedia device, a streaming media device, and the like. The terminal device can be widely applied to various scenarios, such as device-to-device (D2D) communication, vehicle to everything (V2X) communication, machine-type communication (MTC), internet of things (IOT), virtual reality, augmented reality, industrial control, autonomous driving, remote medical treatment, smart power grid, smart furniture, smart office, smart wear, smart transportation, smart city, and the like. The terminal device can be a mobile phone, a tablet computer, a computer with wireless transceiver function, a wearable device, a vehicle, an urban air vehicle (such as a pilotless plane, a helicopter, and the like), a ship, a robot, a mechanical arm, a smart home device, and the like.

[0113] The RAN is configured to provide wireless access for the terminal device, including but not limited to a 5G base station (gNB), a next-generation base station in the 6th generation (6G) mobile communication system, a base station in a future mobile communication system, a wireless base station (eNodeB or eNB) in an LTE network, a wireless fidelity access point (Wi-Fi AP), a worldwide interoperability for microwave access base station (WiMAX BS), a relay station, and the like; or a module or unit that completes part of the function of the base station. In the 5G RAN architecture, the gNB can include a centralized unit (CU) and a distributed unit (DU). The gNB can also include a radio unit (RU). The CU and the DU can be understood as a division of the base station from the perspective of logical functions. The CU and the DU can be separated physically or deployed together. For example, multiple DUs can share one CU or one DU can be connected to multiple CUs. The CU and the DU can be connected through an F1 interface.

[0114] The AMF network element is mainly responsible for mobility management in the mobile network, such as user location update, user registration network, user handover, and the like. In addition, the AMF network element is also responsible for delivering user policies between a terminal device and a PCF. The SMF is mainly responsible for session management in the mobile network, such as session establishment, modification, and release. Specific functions include allocating an internet protocol (IP) address for a user, selecting a UPF that provides message forwarding functions, and the like. The PCF is responsible for providing policies to the AMF and the SMF, such as quality of service (QoS) policies and slice selection policies. The UPF is mainly responsible for processing user messages, such as forwarding, charging, and the like.

[0115] The NWDAF network element has functions of data collection, model training, data analysis, or model inference. The NWDAF network element supports collecting data from other network functions and application function (AF) network elements, supports collecting data from an operation administration and maintenance (OAM) network element, and supports providing analysis information to other network functions and AF network elements.

[0116] The NWDAF network element can be used to collect relevant data from network function network elements, third-party service servers, terminal devices, or network management systems, perform data analysis based on the relevant data to obtain analysis results, and provide the analysis results to the network function network elements, third-party service servers, terminal devices, or network management systems. The analysis results can assist the network in selecting service quality parameters of a service, or assist the network in performing traffic routing, or assist the network in selecting background data transmission strategies, and the like. In addition, the NWDAF network element can also be used to collect relevant data from network function network elements, third-party service servers, terminal devices, or network management systems, and perform model training based on the relevant data to obtain an artificial intelligence (AI) model or a machine learning (ML) model, and provide the AI model / ML model to other NWDAF network elements. The AI model / ML model can be used to assist the NWDAF network element in generating data analysis results based on relevant data.

[0117] 3GPP splits the training function and inference function of the NWDAF, one NWDAF can support only the model training function, or only the data inference function, or both the model training function and the data inference function. Among them, the NWDAF supporting the model training function can also be called a training NWDAF, or a NWDAF containing a model training logical function (MTLF) (referred to as MTLF for short). The training NWDAF can perform model training according to the obtained data to obtain a trained model. The NWDAF supporting the data inference function can also be called an inference NWDAF, or a NWDAF containing an analytics logical function (AnLF) (referred to as AnLF for short). The inference NWDAF can input input data into the trained model to obtain analysis results or inference data. In the embodiments of the present application, the training NWDAF refers to the NWDAF supporting at least the model training function. As a possible implementation manner, the training NWDAF can also support the data inference function. The inference NWDAF refers to the NWDAF supporting at least the data inference function. As a possible implementation manner, the inference NWDAF can also support the model training function. If a NWDAF supports both the model training function and the data inference function, the NWDAF can be called a training NWDAF, an inference NWDAF, or a training and inference NWDAF, or a NWDAF. In the embodiments of the present application, one NWDAF can be a single network element, or can be combined with other network elements, for example, the NWDAF is set to a policy control function (PCF) network element or an access and mobility management function (AMF) network element.

[0118] The NWDAF network element in the present application can simultaneously support the model training function and / or the model (data) inference function.

[0119] The UDM network element can be used to store user data, such as subscription information, authentication or authorization information, policy data, application data, and the like.

[0120] The NRF network element can provide registration and discovery functions, and can enable network functions (network functions, NFs) to discover each other and communicate through application programming interfaces (application programming interfaces, APIs). The NRF network element can also provide other network element management services, such as network element updating, deregistration, and network element state subscription and pushing, and the like.

[0121] The NEF network element can be used to support the opening of capabilities and events. For example, opening the capabilities of each network function (NF), and converting internal and external information. The NEF network element can provide security assurance to ensure the security of external applications to the 3GPP network, provide external application quality of service (QoS) customization capability opening, mobility state event subscription, AF request distribution, and other functions.

[0122] The AF network element can be used to deliver application-side requirements to the network side, such as quality of service (QoS) requirements or user state event subscriptions. The AF can be a third-party functional entity or an application server deployed by an operator, and the embodiments of the present application are not limited.

[0123] It can be understood that the above network elements are examples of an implementation manner, and the present application does not exclude that there are network elements or devices with the above network element functions in the 6G or future wireless communication system, or have other names or other forms.

[0124] It can be understood that each network function shown in FIG. 1 can refer to a related protocol or standard, and the present application does not expand on the description. It can be understood that “Nnef”, “Nnrf”, “Npcf”, “Nudm”, “Nudr”, “Nnwdaf”, “Naf”, “Nausf”, “Namf”, and “Nsmf” in FIG. 1 represent the name of the service interface, which can be used to call the corresponding service operation. For details, see the related description in the 3GPP standard protocol, which is not expanded here.

[0125] It should also be understood that N1, N2, N3, N4, N6, etc. shown in FIG. 1 are interface sequence numbers. The meanings of these interface sequence numbers are as follows: N1 can represent the interface between the AMF network element and the terminal device, and can be used to deliver non-access stratum (NAS) signaling (such as QoS rules from the AMF network element) to the terminal device, etc. N2 can represent the interface between the AMF network element and the RAN, and can be used to deliver core network side to RAN radio bearer control information, etc. N3 can represent the interface between the RAN and the UPF network element, and is mainly used to deliver uplink and downlink user plane data between the RAN and the UPF network element. N4 can represent the interface between the SMF network element and the UPF network element, and can be used to deliver information between the control plane and the user plane, including the delivery of control plane forwarding rules, QoS rules, traffic statistics rules, etc. to the user plane, and the reporting of user plane information. N6 can represent the interface between the UPF network element and the DN, and can be used to deliver uplink and downlink user data flow between the UPF network element and the DN. For example, the meanings of the above interface sequence numbers can also be referred to the meanings defined in the 3GPP standard protocol, and the meanings of the above interface sequence numbers are not limited in the present application.

[0126] In a possible implementation manner, the network element or function in the embodiments of the present application can be a network element in a hardware device, or a software function running on a dedicated hardware, or a virtualized function instantiated on a platform (for example, a cloud platform). As a possible implementation manner, the network element or function in the embodiments of the present application can be implemented by one device, or can be implemented by multiple devices together, or can be a functional module in one device, and the embodiments of the present application do not make a specific limitation in this regard.

[0127] In order to better understand the technical solutions of the embodiments of the present application, some related contents involved in the present application are briefly described below.

[0128] I. Network function (NF) service registration and update

[0129] Referring to FIG. 2, FIG. 2 is a flow diagram of NF service registration provided by the embodiments of the present application. As shown in FIG. 2, the flow of NF service registration includes but is not limited to the following steps:

[0130] Step 1: The NF service consumer (NF instance) sends an NF registration request (for example, Nnrf_NF Management_NF Register_request) to the NRF to inform the NRF of its NF profile when the NF service consumer is first operable. The NF profile of the NF service consumer can be configured by the OAM system.

[0131] Step 2, the NRF stores the NF profile of the NF service consumer and marks the NF service consumer as available. Whether the NF profile sent by the NF service consumer to the NRF needs to be integrity protected by the NF service consumer and verified by the NRF can be decided by the Standalone (SA) SA3.

[0132] Step 3, the NRF confirms the NF registration is accepted through an NF registration response (for example, Nnrf_NF Management_NF Register_response).

[0133] Referring to FIG. 3, FIG. 3 is a flow diagram of NF service update provided by an embodiment of the present application. As shown in FIG. 3, the flow of NF service update includes but is not limited to the following steps:

[0134] Step 1, the NF service consumer, i.e., the NF instance, sends an NF update request (for example, Nnrf_NF Management_NF Update_request) to the NRF, which can carry the NF profile updated by the NF service consumer to inform the NRF of the updated NF profile.

[0135] Step 2, the NRF updates the NF profile of the NF service consumer.

[0136] Step 3, the NRF confirms the NF update is accepted through an NF update response (for example, Nnrf_NF Management_NF Update_response).

[0137] For example, the NF instance ID is included in the NF update request (for example, Nnrf_NF Management_NF Update_request). In a possible implementation, if the complete NF profile is updated, the complete NF profile can be provided in the NF update request (for example, Nnrf_NF Management_NF Update_request); if part of the NF profile is updated, the NF profile element that needs to be updated can be provided in the NF update request (for example, Nnrf_NF Management_NF Update_request).

[0138] It can be understood that the "NF service consumer" in the above-mentioned FIG. 2 and FIG. 3 can refer to the consumer of the NRF service, and should not be confused with the role of the NF (consumer or producer).

[0139] II. Network Function (NF) Discovery

[0140] Referring to FIG. 4, FIG. 4 is a flow diagram of NF discovery / NF service discovery according to an embodiment of the present application. FIG. 4 shows a NF discovery procedure or a NF service discovery procedure for a NF service consumer in the same public land mobile network (PLMN). As shown in FIG. 4, the NF discovery / NF service discovery procedure includes but is not limited to the following steps:

[0141] In step 1, a NF service consumer wants to discover available services in the network based on service name and target NF type. The NF service consumer can send a NF discovery request (e.g. Nnrf_NF Discovery_Request) to the NRF in the same PLMN. The NF discovery request (e.g. Nnrf_NF Discovery_Request) can include but is not limited to: expected NF service name, expected NF type of NF instance, and NF type of NF consumer. For example, the NF discovery request can further include one or more of the following: producer NF set identifier (ID), NF service set ID, subscription permanent identifier (SUPI), data set identifier, external group ID (for UDM, UDR discovery), routing indication of UE and home network public key identifier (for UDM and AUSF discovery), single network slice selection assistance information (S-NSSAI), network slice instance (NSI ID) (if available), and other service related parameters. In addition, for AMF discovery, the NF discovery request can further include AMF area identity, AMF set identity, tracking area identity (TAI). It can be understood that the use of NSI ID within a PLMN depends on network deployment. It can also be understood that the need for other service related parameters depends on the NF type of the expected NF instance, which can be found in section 6.3 of 3GPP TS 23.501 [2] “Principles of network function and network function service discovery and selection”, which is not described in detail here.

[0142] In one possible implementation, the NF service consumer can indicate a preference for a target NF location in the Nnrf_NF Discovery_Request. In this case, the NF service consumer can indicate that its NF location is preferred over the target NF location.

[0143] Step 2, NRF authorizes NF service discovery. NRF determines whether to allow the NF service consumer to discover the desired NF instance(s) based on the attributes of the desired NF / NF service and the type of the NF service consumer. If the desired NF instance(s) or NF service instance(s) are deployed in a certain network slice, NRF authorizes the NF discovery request based on the discovery configuration of the network slice. For example, the discovery configuration of the network slice is that the desired NF instance(s) can only be discovered by the NFs in the same network slice.

[0144] Step 3, if the authorization is allowed, NRF determines the NF instance set that matches the NF discovery request (e.g. Nnrf_NF Discovery_Request) and the NRF internal policy, and downlink the NF profile of the determined NF instance(s). The NF profile of each determined NF instance can output the required parameters to the NF service consumer through the NF discovery request response (e.g. Nnrf_NF Discovery_Request Response).

[0145] In a possible implementation, if the target NF is UDR, UDM or AUSF, and SUPI is used as an optional input parameter in the NF discovery request, NRF provides the corresponding UDR, UDM or AUSF instance(s) that matches the optional input SUPI. Otherwise, NRF can return all applicable UDR instance(s) (e.g. determined based on data set ID, NF type), UDM instance(s) (e.g. determined based on NF type) or AUSF instance(s) (e.g. determined based on NF type). Optionally, NRF can also return the range information of SUPI(s) and / or data set ID supported by each UDR instance.

[0146] In a possible implementation, if the target NF is charging function (CHF), and SUPI, generic public subscription identifier (GPSI) or PLMN ID is used as an optional input parameter in the NF discovery request, NRF provides the corresponding CHF instance(s) that matches the optional input SUPI, GPSI or PLMN ID. If a pair of primary CHF instance and secondary CHF instance is configured in the CHF instance profile, NRF can provide the pair of primary CHF instance and secondary CHF instance at the same time. Otherwise, if the target NF is CHF, but no SUPI / PLMN ID and GPSI are provided in the NF discovery request, NRF can return all applicable CHF instance(s). Optionally, NRF can also return the range information of SUPI(s), GPSI(s) or PLMN ID(s).

[0147] In a possible implementation, if the NF service consumer provides a preferred target NF location, the NRF can not limit the discovered set of NF instances or NF service instances to the target NF location. For example, if no NF instance or NF service instance can be found at the preferred target NF location, the NRF can provide an NF instance or NF service instance whose location is not the preferred target NF location.

[0148] III. Sample alignment procedure with NWDAF as VFL active participant

[0149] Referring to FIG. 5, FIG. 5 is a schematic diagram of a sample alignment procedure according to an embodiment of the present application. As shown in FIG. 5, the sample alignment procedure includes a vertical federated learning (VFL) server (VFL Server) and one or more VFL participants (VFL Participant(s)). The VFL server can guide the training and / or distributed inference process, and the VFL participants follow the instructions of the VFL server. Optionally, the VFL server can be co-located with one or more VFL participants. For example, the VFL server can be co-located with one or more VFL participants on the same physical device or logical device, such as the same network element (e.g., NWDAF, AF, etc.). For example, the NWDAF network element can simultaneously serve as a VFL server and a VFL active participant, or the NWDAF network element can simultaneously serve as a VFL server and a VFL passive participant.

[0150] In a possible implementation, the VFL server (VFL Server) can be an NWDAF or AF that integrates local training results into local ML model updates in the VFL training process. It also coordinates the VFL training process by discovering and selecting VFL clients. In the VFL inference process, the VFL server aggregates local inference results from VFL clients, generates the final VFL inference result, and sends the final VFL inference result to the consumer. There is only one VFL server per VFL process. The VFL client (VFL Client) can be an NWDAF or AF that maintains a local dataset and performs local training and inference according to the requirements of the VFL server. There can be multiple VFL clients in the VFL training and inference.

[0151] Exemplarily, the VFL server in FIG. 5 is a NWDAF, which can contain both a Model Training Logic Function (MTLF) and an Analytics Logic Function (AnLF). It can be understood that the VFL training phase involves the MTLF, and the VFL inference involves the AnLF. If the MTLF and the AnLF are not co-located, the trained model can be shared by the MTLF to the consumer AnLF. In the sample alignment procedure shown in FIG. 5, the VFL server (e.g., the NWDAF) coordinates the VFL operation, and optionally, it can also act as a VFL active participant and can access the labels. In addition, the VFL participants in FIG. 5 can be VFL active participants with label access rights, or VFL passive participants without label access rights. The VFL participants can also be referred to as VFL customers. Exemplarily, the party that can access or provide labels in the VFL task is referred to as a VFL active participant. Correspondingly, the party that does not have access to labels or cannot provide labels in the VFL task is referred to as a VFL passive participant.

[0152] As shown in FIG. 5, the sample alignment procedure can include but is not limited to the following steps:

[0153] Step 1, the VFL server (e.g., the NWDAF) and the VFL participants (e.g., the NWDAF, the AF) register with the NRF. The registered information can include the NF profile, the analytics ID(s), the address information of the NWDAF, the service area, the VFL capability type information (e.g., the VFL server or the VFL participant type), and the time interval that supports the VFL. The value of the VFL participant type can be “active” or “passive”.

[0154] Step 2, VFL server / participant discovery and initial participant selection. The VFL server and the VFL participants are discovered by invoking the network element discovery request (e.g., the Nnrf_NF Discovery_Request) service operation. Exemplarily, the initial selection of the VFL participants by the VFL server can occur in this step. Alternatively, the selection of the VFL participants by the VFL server can also be completed in Step 7 described below.

[0155] Step 3, the VFL server sends a VFL preparation request to the VFL participants. For the NWDAF participant, the existing service operation can be reused and enhanced, such as the machine learning model training information request (e.g., Nnwdaf_ML Model Training Info_Request) or the machine learning model training subscription (e.g., Nnwdaf_ML Model Training_Subscribe), or a new service can be defined. For the AF participant, a new AF / NEF service can be defined. In either way, a ML preparation flag will be provided in the VFL preparation request to check whether the VFL participant is able to meet the ML model training requirements (e.g., analytics ID, ML model interoperability information, available data requirements, available time requirements, etc.). The VFL participant can respond to the VFL server to indicate whether it will join the VFL operation or not, and if the VFL participant is not able to join the VFL operation, the reason why it cannot join the VFL operation can be carried in the response message.

[0156] Step 4, the VFL server can determine the list of required features and target samples and send a data alignment request (i.e., samples and features) to the VFL participants (e.g., AF participants) through the NEF, which includes the information required for the alignment of features and samples. This data alignment request can facilitate the alignment of datasets from different sources in the VFL environment. It can enable the participating entities of the VFL task to use a common set of samples without revealing sensitive data. The samples can be the intersection of different datasets.

[0157] Exemplarily, the above data alignment request includes a dataset identifier (i.e., a unique identifier of the dataset held by each VFL participant), an alignment technique (i.e., a specific method or algorithm to be used for data alignment, such as private set intersection (PSI), feature hashing, or other techniques to ensure data privacy and integrity), and a notification target address. Optionally, the data alignment request can also include an alignment correlation ID, an expiration time, and additional data alignment information, such as challenges or differences encountered in the data alignment process.

[0158] Step 5, each VFL participant performs data alignment pre-processing to determine its capability to meet the AI / ML model training requirements. The data alignment pre-processing can include model training requirement feasibility assessment, model access and compatibility, data alignment verification, etc. In this step, the VFL passive participant can determine the list of supported features and samples. The passive VFL participant can determine the list of supported features (e.g., a subset of the required feature list shared by the VFL active participant) and the list of supported samples (e.g., a subset of the target samples shared by the active participant) based on the information received from the VFL active participant, its available data, request the corresponding NF type or instance of the analysis ID, the latest data collection operation, and the ML model interoperability information.

[0159] Step 6, the VFL participants (e.g., AF participants) inform the VFL server of the results of the data alignment through the NEF. In addition, the VFL participants can provide the VFL server with decisions on their participation in the VFL operation and the identified feature and sample IDs (if any). If the VFL participant cannot join the VFL operation, it can also include the reason why it cannot join the VFL operation in the response message.

[0160] Step 7, the VFL server selects the VFL participants based on the inputs received from the VFL participants. In this step, the VFL active participant can perform sample alignment by identifying the overlap / crossover of the samples supported by all VFL passive participants. If the list of samples supported by the VFL passive participant selected in Step 4 has no or little overlap with the samples supported by other VFL passive participants, it can be excluded from the VFL training.

[0161] Based on the features supported by each VFL passive participant (feature dimensions of each VFL passive participant), the VFL active participant can also decide how to partition the data features among the VFL passive participants and assign a subset of features to each VFL passive participant. If the list of features supported by the selected VFL passive participant in step 4 has no overlap with the list of required features, it can be excluded from the VFL training. Since the analytics ID specified by 3GPP requires input data from different domains including 5GC (5G core network) and AF, domain-based feature selection for cross-domain VFL can be used, where each domain performs VFL training based on the data owned by that domain (i.e., AF performs VFL training based on AF data, while 5GC performs VFL training based on 5GC data). However, in each domain, it is important to decide how to vertically partition the data features among multiple VFL participants. For example, within the 5GC domain, the observed traffic experience analytics ID requires input data from AMF, SMF, and UPF. In one possible approach, the VFL passive participants can perform VFL training together based on AMF, SMF, and UPF data. In another possible approach, each VFL passive participant can perform VFL training based on different NF types.

[0162] Step 8 is the VFL training procedure, which is not elaborated here.

[0163] Step 9, the Consumer NF sends an analytics subscription / request to the VFL server for requesting VFL inference.

[0164] Step 10 is the VFL inference procedure, which is not elaborated here.

[0165] Step 11, the VFL server returns an analytics notification to the Consumer NF, which includes the VFL inference result.

[0166] Four, sample alignment procedure with AF as active participant

[0167] Referring to FIG. 6, FIG. 6 is a schematic diagram of another sample alignment procedure provided by the embodiments of the present application. As shown in FIG. 6, the sample alignment procedure includes an active participant and passive participant(s). For example, the active participant in FIG. 6 is AF, and the passive participant is NWDAF. The NWDAF can include a model training logic function (MTLF) and an analytics logic function (AnLF) at the same time. It can be understood that the VFL training phase involves the MTLF, and the VFL inference involves the AnLF. If the MTLF and the AnLF are not co-located, the trained model can be shared by the MTLF to the consumer AnLF.

[0168] As shown in FIG. 6, the sample alignment procedure can include but is not limited to the following steps:

[0169] Step 1, the VFL active participant (i.e., VFL server) (such as AF) decides to use the VFL model training. Because the data cannot be directly obtained / exposed from the 5GC (for example, for privacy reasons) or the vendor implementation policy, the VFL active participant (i.e., VFL server) (such as AF) sends a VFL participant discovery request to the NEF by indicating the analytics ID (for example, observed service experience), application ID, external UE ID, filter information (for example, area of interest).

[0170] Step 2, based on the request from the AF, the NEF authorizes the request information and applies parameter mapping (for example, external UE ID mapping to internal UE ID, geographic area mapping to TA(s) / Cell-id(s), VFL passive participant (VFL client) capability). The NEF requests the VFL passive participant (VFL client) discovery from the NRF, and if the request information provided by the AF supports the NWDAF registration in the UDM, the NEF further queries the UDM to determine the NWDAF ID and its corresponding service UE list. The service UE list of each NWDAF can be a subset of the UE requested by the AF.

[0171] Step 3, the NRF returns the available NWDAF(s) supporting the VFL passive participant (VFL client) capability and the requested analytics ID to the NEF through the candidate passive participant response. The UDM returns the available NWDAF(s) and its corresponding service UE list to the NEF.

[0172] Step 4, the NEF sends a VFL participant discovery response to the AF, sending the available NWDAF(s) and their corresponding list of served UEs to the AF.

[0173] Step 5, the VFL active participant (i.e. VFL server) (e.g. AF) analyses the ID (e.g. observed service experience) based on the information received in step 4, and sends a VFL sample and feature alignment request to the NEF. The VFL sample and feature alignment request includes the sample (i.e. external UE ID) and Application ID that the AF wants to use for VFL training, where the Application ID indicates which application needs to initiate the VFL operation. Optionally, the VFL sample and feature alignment request also includes a feature profile, indicating which features the AF can provide and which features are needed to perform VFL. For example, the AF can indicate in the feature profile that for the observed service experience case, the access speed, stall time, frame rate of the AF can be used as features.

[0174] Step 6, the NEF maps the external UE ID to an internal UE ID based on the information interacted with the UDM.

[0175] Step 7, the NEF sends a VFL sample and feature alignment request to the VFL passive participant (i.e. VFL client) (e.g. NWDAF). The VFL sample and feature alignment request includes the analysis ID, internal UE ID (e.g. SUPI), Application ID and feature profile. If multiple NWDAFs are involved, the NEF sends the VFL sample and feature alignment request to each NWDAF separately.

[0176] Step 8, based on the feature profile and requested analysis ID, the NWDAF(s) collects input data related to the internal UE ID, as described in TS 23.288 clause 6.4.2.

[0177] Step 9, based on the input data collected from the NFs, the NWDAF(s) can determine which UE related data is available, whether the features of the input data can fulfil / satisfy the feature profile provided by the AF. The NWDAF can select the UEs from the internal UE ID to guarantee the same sample between the NWDAF(s) and the AF. On the other hand, the features in the same sample can be different between the NWDAF and the AF.

[0178] Step 10, NWDAF(s) sends VFL sample and feature alignment response to NEF, which includes application ID, sample and feature alignment indication. The sample and feature alignment indication can be used to indicate which UE and which feature can be used for VFL operation.

[0179] Step 11, NEF maps internal UE ID to external UE ID. If multiple NWDAFs are involved, NEF collects VFL sample and feature alignment response from NWDAFs. The UE ID in this step can be a subset of UE ID in AF request based on NWDAF(s) determination for sample and feature alignment.

[0180] Step 12, NEF sends VFL sample and feature alignment response to AF, including sample and feature alignment indication and application ID.

[0181] Step 13, VFL active participant (i.e. VFL server) verifies whether the selected sample and feature meet the requirements. If the VFL active participant (i.e. VFL server) determines that the number of samples or features is insufficient, it can stop the VFL process. Subsequently, it has the ability to start a new VFL process by adjusting the requirements or selecting different passive participant entities.

[0182] Step 14, when a new VFL passive participant (V i.e. FL client) is introduced, or when the current VFL active participant (i.e. VFL server) needs to be updated, the VFL active participant (i.e. VFL server) assesses whether a realignment of the ongoing VFL process is needed, or whether a realignment for starting a new collaborative learning process is needed. If needed, the initial sample and feature alignment procedure is repeated. The realignment can ensure that the sample and feature requirements are properly matched to facilitate effective collaboration. The VFL active participant (i.e. VFL server) can also start the sample and feature alignment process according to the existing alignment requirements or updated version.

[0183] Vertical federated learning (VFL) tasks require each VFL participant to use sample data of the same object for training. For example, VFL participants (including: NWDAF, AF) 1, 2, 3, each VFL participant uses the data of UE1, UE2, UE3 for vertical federated learning training. However, the types of sample identifiers stored in different AFs can not be the same, for example, AF1 uses IP address (IP address) to mark user 1, and AF2 uses GPSI to mark user 1. At this time, AF1 and AF2 will cause sample identifier mismatch when performing sample alignment, so that sample alignment cannot be achieved.

[0184] Based on this, the application provides a sample alignment method, device and readable storage medium in a longitudinal federated learning, which can reduce the possibility of sample identification mismatch in the sample alignment process of the longitudinal federated learning task, and realize sample alignment in the longitudinal federated learning task.

[0185] The technical solutions provided by the application will be described in detail below in combination with more drawings.

[0186] The technical solutions provided by the application are described through multiple embodiments, and specific reference is made to the description of each embodiment below. In the application, the same or similar parts of each embodiment or implementation can be mutually referenced. In the application, each embodiment and each implementation / implementation method / realization method in each embodiment, if not specially stated and logically conflicted, the terms and / or descriptions of different embodiments and each implementation / implementation method / realization method in each embodiment are consistent and can be mutually referenced, and the technical features of different embodiments and each implementation / implementation method / realization method in each embodiment can be combined to form new embodiments, implementations, implementation methods or realization methods according to their inherent logical relationship. The implementation methods of the application described below do not constitute a limitation on the protection scope of the application.

[0187] In the application, "network element A sends information A to network element B" can be understood as that the destination of the information A or the intermediate network element in the transmission path between the destination is network element B, which can include direct or indirect sending of information to network element B. "Network element B receives information A from network element A" can be understood as that the source of the information A or the intermediate network element in the transmission path between the source is network element A, which can include direct or indirect receiving of information from network element A. The information can be processed as necessary between the source and the destination of the information sending, such as format change, etc., but the destination can understand the valid information from the source. Similar expressions in the application can be understood similarly, which will not be described here.

[0188] It should be understood that in the application, the indication includes direct indication (also known as explicit indication) and implicit indication. Among them, the direct indication information A means including the information A; the implicit indication information A means indicating the information A through the corresponding relationship between the information A and the information B and the direct indication information B. Among them, the corresponding relationship between the information A and the information B can be predefined, pre-stored, pre-burned or pre-configured.

[0189] It should be understood that in the application, determining information D based on information C includes determining information D based only on information C, and determining information D based on information C and other information. In addition, information C used to determine information D can also include indirect determination, such as information D is determined based on information E, and information E is determined based on information C.

[0190] The various embodiments will be described in detail below.

[0191] Referring to FIG. 7, FIG. 7 is a flow diagram of a sample alignment method in vertical federated learning according to an embodiment of the present disclosure. In this method, the first network element can be NWDAF, which can act as a VFL active participant or a VFL server; the second network element can be AF, which can act as a VFL passive participant or a VFL client. The second network element in this method can have one or more.

[0192] As shown in FIG. 7, the sample alignment method in vertical federated learning includes but is not limited to the following steps:

[0193] S101, the first network element (NWDAF) and the second network element (AF) register their capabilities to the NRF network element respectively.

[0194] In one possible implementation, the first network element (NWDAF) can send a network element registration request (for example: Nnrf_NF Management_NF Register_request) to the NRF network element for registering its capability. The network element registration request can include the identification type of the sample supported by the first network element (NWDAF). For example, the network element registration request can also include one or more of the following: analysis ID, VFL task group (such as AF ID, Vendor ID (vendor identification) or UE ID, etc.) that the first network element (NWDAF) joins, or role attribute (such as participant (whether it can provide a label), coordinator) supported by the first network element (NWDAF). It can be understood that the role attribute here is for the VFL task group or the analysis ID. Accordingly, the NRF network element returns a network element registration response (for example: Nnrf_NF Management_NF Register_response) to the first network element (NWDAF) to confirm that the registration of the first network element (NWDAF) is accepted.

[0195] Similarly, one or more second network elements (AFs) send a network element registration request (e.g., Nnrf_NFManagement_NFRegister_request) to the NRF network element, respectively, for registering their capabilities. The network element registration request can include the identification type of the samples supported by the second network element (AF). Exemplarily, the network element registration request can further include one or more of the following: an analytics ID, a VFL task group that the second network element (AF) joins (e.g., an AF ID, a Vendor ID (vendor identification), or a UE ID, etc.), or a role attribute (e.g., a participant (whether it can provide a label), a coordinator) supported by the second network element (AF). The role attribute is for the VFL task group or the analytics ID. Correspondingly, the NRF network element returns a network element registration response (e.g., Nnrf_NFManagement_NFRegister_response) to the second network element (AF) for confirming that the registration of the second network element (AF) is accepted.

[0196] In another possible implementation, one or more second network elements (AFs) can send a network element registration request 1 to the NEF network element, respectively, for registering their capabilities. The network element registration request 1 can include the identification type of the samples supported by the second network element (AF). Exemplarily, the network element registration request 1 can further include one or more of the following: an analytics ID, a VFL task group that the second network element (AF) joins (e.g., an AF ID, a Vendor ID (vendor identification), or a UE ID, etc.), or a role attribute (e.g., a participant (whether it can provide a label), a coordinator) supported by the second network element (AF). The role attribute is for the VFL task group or the analytics ID. After receiving the network element registration request 1, the NEF network element can send a network element registration request 2 to the NRF network element for registering the capabilities of the NEF network element. The network element registration request 2 can include the profile of the NEF network element, the second network elements (AFs) supported by the NEF network element, the identification type of the samples supported by the second network elements (AFs), etc. The NRF network element returns a network element registration response 2 to the NEF network element for confirming that the registration of the NEF network element is accepted. The NEF network element returns a network element registration response 1 to the second network element (AF) for confirming that the registration of the second network element (AF) is accepted. In this implementation, the AF is registered in the NRF network element as a part of the NEF network element, and in the subsequent network element discovery process, the NRF network element can discover the AF through the NEF network element.

[0197] In the embodiments of the present application, the role attribute supported by the first network element (NWDAF) is a coordinator (i.e., a VFL server) or a VFL active participant, and the role attribute supported by the second network element (AF) is a participant (without providing a label) or a VFL passive participant (also referred to as a VFL slave participant).

[0198] In a possible implementation, the sample supported by one network element can be understood as a sample that can be used by the network element for machine learning, artificial intelligence tasks (training and / or inference), for example, a vertical federated training task.

[0199] In a possible implementation, the identification type of the sample supported by one network element can be one or more. For example: SUPI, GPSI, or UE IP address, etc. Wherein the UE IP address can be one or more of the following: IPV4, IPV4 and port number, IPV6, or IPV6 prefix (IPV6 prefix).

[0200] It can be understood that one sample supported by a network element in the embodiment of the present application can include multiple types of data, and the identification of one sample can correspond to multiple data types, such as the identification 1 of the sample corresponding to the data type 1, the data type 2, and the data type 3, as shown in Table 1 below, the data types corresponding to the identifications of different samples can not be completely the same (partially the same or completely different), of course, they can also be the same, and the embodiment of the present application is not limited. The sample identified by the identification of one sample includes (corresponds to) data corresponding to multiple data types. It can also be understood that the identification type of the sample and the type of the data are not the same in the embodiment of the present application.

[0201] Table 1

[0202] In a possible implementation, the identification type of the sample can be the granularity of the analysis ID, or the granularity of the VFL group, and the embodiment of the present application is not limited. In other words, for the network element with the same analysis ID, the identification type of the sample supported by the network element is the same; for the members in the same VFL group, the identification type of the sample supported by the members has an intersection.

[0203] S102, the first network element (NWDAF) sends a network element discovery request to the NRF network element. Correspondingly, the NRF network element receives the network element discovery request.

[0204] S103, the NRF network element sends a network element discovery response to the first network element (NWDAF), and the network element discovery response includes the information of the second network element.

[0205] Correspondingly, the first network element (NWDAF) receives the network element discovery response.

[0206] S104, the first network element (NWDAF) determines the identification type of the sample used in the vertical federated learning task. Wherein the identification type of the sample used in the vertical federated learning task includes one or more in the intersection between the identification type of the sample supported by the first network element (NWDAF) and the identification type of the sample supported by the second network element (AF).

[0207] In a possible implementation, the first network element (NWDAF) as a VFL active participant can send a network element discovery request (e.g., Nnrf_NFDiscovery_Request) to the NRF network element for discovering or screening VFL passive participants (e.g., AFs). The network element discovery request can include but is not limited to an expected NF service name, an NF type of an expected NF instance, or an NF type of an NF consumer, etc. The content included in the network element discovery request can refer to the prior art, which is not described here. After receiving the network element discovery request, the NRF network element can select one or more second network elements (AFs) based on the information carried in the network element discovery request, and can return a network element discovery response (e.g., Nnrf_NFDiscovery_Request Response) to the first network element (NWDAF). The network element discovery response can include information of the one or more second network elements (AFs), for example, an identification type of samples supported by the second network element (AF), and optionally, an identification of the second network element (AF). After receiving the network element discovery response, the first network element (NWDAF) can determine an identification type of samples used in a vertical federated learning (VFL) task based on an identification type of samples supported by the first network element (NWDAF) and the identification type of samples supported by the second network element (AF) in the network element discovery response. For example, the identification type of samples used in the VFL task can include one or more of an intersection between the identification type of samples supported by the first network element (NWDAF) and the identification type of samples supported by the second network element (AF).

[0208] In a possible implementation, the network element discovery request described above can further include one or more identification types of samples supported by the first network element (NWDAF). For example, the first network element (NWDAF) can first determine an identification type (SUPI, GPSI, or IP address) of samples supported by the first network element (NWDAF), and can carry the one or more identification types of samples supported by the first network element (NWDAF) in the network element discovery request and send to the NRF network element. At this time, after receiving the network element discovery request, the NRF network element can screen one or more second network elements (AFs) supporting the one or more identification types from the locally stored information, and can return a network element discovery response to the first network element (NWDAF). The network element discovery response can include information of the second network element (AF), for example, an identification of the second network element (AF). In this case, after receiving the network element discovery response, the first network element (NWDAF) can determine an identification type of samples used in a VFL task from the identification type of samples supported by the first network element (NWDAF). For example, the identification type of samples used in the VFL task is a subset of the identification type of samples supported by the first network element (NWDAF).

[0209] Alternatively, after receiving the network element discovery request, the NRF network element can select one or more second network elements (AFs) based on the information carried in the network element discovery request, the intersection between the identification types of the samples supported by the second network elements (AFs) and the identification types of the samples supported by the first network element (NWDAF) is not empty, and the network element discovery response can be returned to the first network element (NWDAF). The network element discovery response can include information of the second network element (AF), such as the intersection between the identification types of the samples supported by the second network element (AF) and the identification types of the samples supported by the first network element (NWDAF), and optionally the identification of the second network element (AF). After receiving the network element discovery response, the first network element (NWDAF) can use the intersection in the network element discovery response as the identification types of the samples used in the VFL task.

[0210] In another possible implementation, the network element discovery request described above can also include the identification types of the expected supported samples. Here, the identification types of the expected supported samples can be part or all of the identification types of the samples supported by the first network element (NWDAF), or the identification types of the samples not supported by the first network element (NWDAF), which is not limited by the embodiments of the present application. The first network element (NWDAF) can send the identification types of the expected supported samples to the NRF network element in the network element discovery request. At this time, after receiving the network element discovery request, the NRF network element can filter one or more second network elements (AFs) supporting the identification types of the expected supported samples from the locally stored information, and can return the network element discovery response to the first network element (NWDAF). The network element discovery response can include information of the second network element (AF), such as the identification of the second network element (AF). In this case, the first network element (NWDAF) can use the identification types of the expected supported samples as the identification types of the samples used in the VFL task.

[0211] In the present application, the "identification types of the samples used in the VFL task" can be understood as "identification types of the samples used in the sample alignment process of the VFL task". In the subsequent training process of the VFL task, the identification types of the samples can be updated. In other words, the identification types of the samples used in the sample alignment process of the VFL task can be different from the identification types of the samples used in the subsequent training process, of course, they can also be the same, which is not limited by the embodiments of the present application.

[0212] In a possible implementation, after the first network element (NWDAF) determines the identification type of the sample used in the VFL task, the first network element (NWDAF) can determine whether the locally stored sample identification a matches the identification type of the sample used in the VFL task. If the locally stored sample identification a of the first network element (NWDAF) does not match the identification type of the sample used in the VFL task, the first network element (NWDAF) can convert the sample identification a to sample identification b. The sample identification b matches the identification type of the sample used in the VFL task. It can be understood that if the locally stored sample identification a of the first network element (NWDAF) matches the identification type of the sample used in the VFL task, no conversion is needed.

[0213] In a possible implementation, the first network element (NWDAF) can also determine encryption information, use the encryption information to encrypt the sample identification b, to obtain the encrypted identification X of the sample supported by the first network element (NWDAF), for subsequent use. Alternatively, when the locally stored sample identification a of the first network element (NWDAF) matches the identification type of the sample used in the VFL task, the first network element (NWDAF) can determine the encryption information, and can use the encryption information to encrypt the sample identification a, to obtain the encrypted identification X of the sample supported by the first network element (NWDAF), for subsequent use. The encryption information can include an encryption algorithm type, and / or specific information of the encryption algorithm (such as a public key, encryption), and the like. The encryption information can be preconfigured or predefined or specified by a standard, or can be determined in advance by negotiation between the first network element (NWDAF) and the second network element (AF), and any manner in which the first network element (NWDAF) and the second network element (AF) can learn the encryption information is within the protection scope of the present application.

[0214] It can be understood that the encryption algorithm in the present application can be homomorphic encryption (Homomorphic Encryption). Homomorphic encryption refers to that after original data is subjected to homomorphic encryption, a specific operation is performed on the obtained ciphertext, and then the plaintext obtained by performing homomorphic decryption on the calculation result is equivalent to the result obtained by directly performing the same operation on the original data (plaintext).

[0215] For example, assuming that the identification type of the sample used in the VFL task is GPSI, and the sample identification a stored locally by the first network element (NWDAF) is SUPI, the first network element (NWDAF) can request the UDM network element to complete the conversion of SUPI and GPSI. In one possible implementation, the first network element (NWDAF) can send a sample identification mapping request to the UDM network element, where the sample identification mapping request includes the sample identification a (such as SUPI), the identification type of the sample used in the VFL task (such as GPSI), and the identification (AF ID) or application identification (APP Name) of the second network element (AF, which can be one or more). After receiving the sample identification mapping request, the UDM network element can determine the sample identification b (such as GPSI) corresponding to the sample identification a (such as SUPI) in the second network element (AF). Specifically, the UDM network element can determine the sample identification b (such as GPSI) according to the subscription data corresponding to the sample identification a (such as SUPI). This is because the subscription data stores the GPSI corresponding to the SUPI corresponding to the AF ID or App Name. It can be understood that in different second network elements (AF), the same UE can be identified by different GPSIs. Therefore, if the sample identification mapping request includes the identification of multiple second network elements (AF) or multiple application identifications (APP Name), the UDM network element determines multiple sample identifications b (GPSI). The UDM network element can send a sample identification mapping response to the first network element (NWDAF), where the sample identification mapping response can include the sample identification b (such as GPSI) corresponding to the sample identification a (such as SUPI) in the second network element (AF). Where the sample identification a (such as SUPI) corresponds to one sample identification b (such as GPSI) in one second network element (AF). For example, if the first network element (NWDAF) requests to obtain the GPSI corresponding to the SUPI in multiple AFs, the sample identification mapping response can include a list of (SUPI-GPSI) tuples, a list of (AF ID-SUPI-GPSI) tuples, or a list of [AF ID-(SUPI-GPSI)] tuples, and the embodiments of the present application do not limit this.

[0216] S105, the first network element (NWDAF) sends a first request to the second network element (AF), where the first request includes the identification type of the sample used in the vertical federated learning task.

[0217] Correspondingly, the second network element (AF) receives the first request.

[0218] S106, the second network element (AF) sends a first response to the first network element (NWDAF), and the first response includes an identification of a sample supported by the second network element. The identification of the sample supported by the second network element can be encrypted or unencrypted (such as plaintext).

[0219] Correspondingly, the first network element (NWDAF) receives the first response.

[0220] In a possible implementation, after the first network element (NWDAF) determines the identification type of the sample used in the VFL task, the first network element (NWDAF) can send a first request to one or more second network elements (AFs), and the first request can include the identification type of the sample used in the VFL task. The first request can be used to request the sample. The first request can also be referred to as a sample alignment request, a data alignment request, or the like, and the embodiments of the present application do not limit the specific name of the first request.

[0221] Correspondingly, after each second network element (AF) receives the first request, the second network element (AF) can determine whether the locally stored sample identification c matches the identification type of the sample used in the VFL task carried in the first request. If the locally stored sample identification c of the second network element (AF) does not match the identification type of the sample used in the VFL task, the second network element (AF) can convert the sample identification c to sample identification d, which matches the identification type of the sample used in the VFL task. For example, assuming that the identification type of the sample used in the VFL task is GPSI, and the locally stored sample identification c of the second network element (AF) is IP address, the second network element (AF) can map the local IP address to GPSI. If the locally stored sample identification c of the second network element (AF) matches the identification type of the sample used in the VFL task, no conversion is needed. The second network element (AF) can then send a first response to the first network element (NWDAF), and the first response can include an unencrypted identification of a sample supported by the second network element, such as sample identification c or sample identification d.

[0222] In a possible implementation, the second network element (AF) can also determine encryption information, and can use the encryption information to encrypt the sample identifier d to obtain an encrypted identifier Y of the sample supported by the second network element (AF). Alternatively, when the sample identifier c stored locally by the second network element (AF) matches the type of the identifier of the sample used in the VFL task, the second network element (AF) can determine encryption information, and can use the encryption information to encrypt the sample identifier c to obtain an encrypted identifier Y of the sample supported by the second network element (AF). The encryption information can include an encryption algorithm type and / or specific information of the encryption algorithm (for example, a public key, encryption, and the like). The encryption algorithm can be a homomorphic encryption algorithm. The second network element (AF) can further send a first response to the first network element (NWDAF), and the first response can include the encrypted identifier Y of the sample supported by the second network element.

[0223] For example, the first response can also be referred to as a sample alignment response, a data alignment response, or the like, and the application does not limit the specific name of the first response.

[0224] In a possible implementation, the first request further includes: an encrypted first network element (NWDAF) supported sample identifier X. After receiving the first request, each second network element (AF) can determine whether the locally stored sample identifier c matches the identifier type of the sample used in the VFL task carried in the first request. If the locally stored sample identifier c of the second network element (AF) does not match the identifier type of the sample used in the VFL task, the second network element (AF) can convert the sample identifier c to a sample identifier d that matches the identifier type of the sample used in the VFL task. For example, assuming that the identifier type of the sample used in the VFL task is GPSI, and the locally stored sample identifier c of the second network element (AF) is an IP address, the second network element (AF) can map the local IP address to GPSI. The second network element (AF) can determine encryption information, and can use the encryption information to encrypt the sample identifier d to obtain an encrypted second network element (AF) supported sample identifier Y. If the locally stored sample identifier c of the second network element (AF) matches the identifier type of the sample used in the VFL task, the second network element (AF) can determine encryption information, and can use the encryption information to encrypt the sample identifier c to obtain an encrypted second network element (AF) supported sample identifier Y. The encryption information can include an encryption algorithm type and / or specific information of the encryption algorithm (such as a public key, encryption, and the like). The encryption algorithm can be homomorphic encryption. The second network element (AF) can further determine the intersection between the encrypted first network element (NWDAF) supported sample identifier X and the encrypted second network element (AF) supported sample identifier Y in the first request, and can send a first response to the first network element (NWDAF). The first response can include the intersection between the encrypted first network element (NWDAF) supported sample identifier X and the encrypted second network element (AF) supported sample identifier Y.

[0225] In a possible implementation, the encryption information can be preconfigured, predefined, or specified by a standard, or can be determined in advance by the first network element (NWDAF) and the second network element (AF). In another possible implementation, the encryption information can also be carried in the first request. In other words, the first request can further include the encryption information. Any manner that enables the first network element (NWDAF) and the second network element (AF) to learn the encryption information is within the protection scope of the present application. It can be understood that the encryption information used by the second network element (AF) for encryption processing is the same as the encryption information used by the first network element (NWDAF) for encryption processing.

[0226] It can be understood that the first network element and the second network element of the embodiments of the present application encrypt the identifiers of the samples supported by each other, and the identifiers of the samples before encryption are matched with the identifier types of the samples in the VFL task. Therefore, even if the other party cannot or does not want to obtain the plaintext, the encrypted ciphertext can be used to obtain the identifiers of the samples supported by the first network element and the second network element. Not only can the security be improved (achieved by encryption), but also the possibility of sample identifier mismatch in the sample alignment process of the vertical federated learning task can be reduced (this effect is achieved by aligning the identifier types of the samples before encryption), thereby achieving sample alignment in the vertical federated learning task.

[0227] In an optional embodiment, the identifier type of the sample used in the VFL task included in the first request can also be preconfigured in the first network element (NWDAF). It can be understood that if the identifier type of the sample used in the VFL task is preconfigured / predefined / pre-set in the first network element (NWDAF), the steps S101 to S104 can not be executed.

[0228] S107, the first network element (NWDAF) determines the identifiers of the samples supported by the vertical federated learning task group based on the identifiers of the samples supported by the second network element (AF). The vertical federated learning task group includes the first network element (NWDAF) and the second network element (AF).

[0229] In a possible implementation, after the first network element (NWDAF) receives the first response returned by one or more second network elements (AF), the first network element (NWDAF) can determine the identifiers of the samples supported by the VFL task group based on the information in the first response. The VFL task group includes the first network element (NWDAF) and the one or more second network elements (AF). One second network element (AF) returns one first response.

[0230] In a possible implementation, the first response includes an encrypted identification Y of the sample supported by the second network element (AF). The first network element (NWDAF) determines the identification of the sample supported by the VFL task group based on the encrypted identification X of the sample supported by the first network element (NWDAF) and the encrypted identification Y of the sample supported by the second network element (AF) in the first response. The identification of the sample supported by the VFL task group can be an intersection between the encrypted identification X of the sample supported by the first network element (NWDAF) and the encrypted identification Y of the sample supported by the second network element (AF). It can be understood that if the second network element (AF) is M (M is an integer greater than or equal to 2), the first network element (NWDAF) can first determine the intersection between the encrypted identification X of the sample supported by the first network element (NWDAF) and the encrypted identification Y of the sample supported by the second network element (AF) in each first response, thereby obtaining M intersections. The first network element (NWDAF) can then determine the identification of the sample supported by the VFL task group according to the clear identification corresponding to the M intersections.

[0231] For example, M is equal to 2, the NWDAF can first determine an intersection 1 between the encrypted identification X1 of the sample supported by the NWDAF and the encrypted identification Y1 of the sample supported by the AF1, and an intersection 2 between the encrypted identification X2 of the sample supported by the NWDAF and the encrypted identification Y2 of the sample supported by the AF1. Because the sample identification a (such as SUPI) stored locally by the NWDAF corresponds to one sample identification b (such as GPSI) in one AF, the encrypted identification of the sample supported by the NWDAF can be different for different AFs. Here, the encrypted identification X1 of the sample supported by the NWDAF is obtained by encrypting the sample identification b (such as GPSI 1) corresponding to the sample identification a (such as SUPI) in the AF1 by the NWDAF. Similarly, the encrypted identification X2 of the sample supported by the NWDAF is obtained by encrypting the sample identification b (such as GPSI 2) corresponding to the sample identification a (such as SUPI) in the AF1 by the NWDAF. The NWDAF can then determine the identification of the sample supported by the VFL task group according to the clear identification corresponding to the intersection 1 and the clear identification corresponding to the intersection 2.

[0232] For example, M equals 2, for AF1, the GPSI of UE1 in NWDAF is encrypted to get M1, the GPSI of UE2 in NWDAF is encrypted to get M2; for AF2, the GPSI of UE1 in NWDAF is encrypted to get A1, the GPSI of UE2 in NWDAF is encrypted to get A2. Assuming that the first response returned by AF1 includes the encrypted GPSI: M1, M2, and M3, the first response returned by AF2 includes the encrypted GPSI: A1, A2, and A3. Then for AF1, the NWDAF determines the intersection {M1, M2} between the encrypted self-supported sample identity {M1, M2} and the encrypted AF1-supported sample identity {M1, M2, M3} in the first response, the intersection {M1, M2} corresponds to the clear-text identity of UE1 and UE2, in other words, the sample supported by the NWDAF and AF1 jointly is UE1 and UE2. Similarly, for AF2, the NWDAF determines the intersection {A1, A2} between the encrypted self-supported sample identity {A1, A2} and the encrypted AF2-supported sample identity {A1, A2, A3} in the first response, the intersection {A1, A2} corresponds to the clear-text identity of UE1 and UE2, in other words, the sample supported by the NWDAF and AF2 jointly is UE1 and UE2. Therefore, the sample identity supported by the VFL task group jointly is the identity of UE1 and UE2.

[0233] In another possible implementation, the first response includes the intersection between the encrypted first network element (NWDAF)-supported sample identity X and the encrypted second network element (AF)-supported sample identity Y. After receiving the first response, the first network element (NWDAF) can take the intersection between the encrypted first network element (NWDAF)-supported sample identity X and the encrypted second network element (AF)-supported sample identity Y in the first response as the sample identity supported by the VFL task group jointly. It can be understood that if there are M (M is an integer greater than or equal to 2) second network elements (AFs), after receiving M first responses returned by the M second network elements (AFs), the first network element (NWDAF) can take the intersection between the clear-text identities corresponding to the sample identities (ciphertexts) contained in the M intersections in the M first responses as the sample identity supported by the VFL task group jointly.

[0234] In another possible implementation, the first response includes the unencrypted second network element-supported sample identity, such as sample identity c or sample identity d. After receiving the first response, the first network element (NWDAF) can take the intersection between the unencrypted second network element-supported sample identity in the first response and the unencrypted first network element-supported sample identity as the sample identity supported by the VFL task group jointly.

[0235] In a possible implementation, after the first network element (NWDAF) determines the identification of the samples commonly supported by the VFL task group, the subsequent VFL training process can be performed, which is described in detail in the prior art and will not be described here.

[0236] The NWDAF in the embodiment of the present application sends a first request to the AF, which contains the identification type of the samples used in the VFL task, and optionally contains the encrypted identification of the samples supported by the NWDAF, which matches the identification type of the samples used in the VFL task; the AF returns the sample identification matching the identification type of the samples used in the VFL task to the NWDAF, and the sample identification is encrypted; or the AF returns the intersection between the encrypted identification of the samples supported by the NWDAF and the encrypted identification of the samples supported by the AF to the NWDAF, and the identification of the samples supported by the AF matches the identification type of the samples used in the VFL task; the NWDAF determines the identification of the samples commonly supported by the VFL task group (including the NWDAF and the AF) according to the information returned by the AF. Because the identification of the samples carried in the first request by the NWDAF and the identification of the samples returned by the AF both match the identification type of the samples used in the VFL task, and both are encrypted. Therefore, even if the plaintext cannot be obtained or is not obtained, the intersection can be obtained by using the encrypted ciphertext, and the identification of the samples commonly supported by the NWDAF and the AF can be obtained. Not only the security can be improved, but also the possibility of sample identification mismatch in the sample alignment process of the vertical federated learning task can be reduced, thereby realizing the sample alignment in the vertical federated learning task.

[0237] Referring to FIG. 8, FIG. 8 is another flowchart of a sample alignment method in a vertical federated learning provided by an embodiment of the present application. In the method, the first network element can be an AF, which can act as a VFL active participant or a VFL server; the second network element can be a NWDAF, which can act as a VFL passive participant or a VFL client. The second network element in the method can be one or more.

[0238] As shown in FIG. 8, the sample alignment method in the vertical federated learning includes but is not limited to the following steps:

[0239] S201, the first network element (AF) and the second network element (NWDAF) register their capabilities to the NRF network element respectively.

[0240] In a possible implementation, the implementation of step S201 in the embodiment of the present application can refer to the implementation of step S101 of the embodiment shown in the foregoing FIG. 7, which will not be described here.

[0241] S202, the first network element (AF) sends a network element discovery request to the NRF network element. Correspondingly, the NRF network element receives the network element discovery request.

[0242] S203, the NRF network element sends a network element discovery response to the first network element (AF), and the network element discovery response includes information of the second network element.

[0243] Correspondingly, the first network element (AF) receives the network element discovery response.

[0244] S204, the first network element (AF) determines the identification type of the sample used in the vertical federated learning task. The identification type of the sample used in the vertical federated learning task includes one or more in the intersection between the identification type of the sample supported by the first network element (NWDAF) and the identification type of the sample supported by the second network element (AF).

[0245] In a possible implementation manner, the implementation manners of steps S202 to S204 in the embodiments of the present application can refer to the implementation manners of steps S102 to S104 in the embodiment shown in FIG. 7, which will not be described herein.

[0246] S205, the first network element (AF) sends a first request to the second network element (NWDAF) through the NEF network element, and the first request includes the identification type of the sample used in the vertical federated learning task.

[0247] Correspondingly, the second network element (NWDAF) receives the first request.

[0248] In a possible implementation manner, the first network element (AF) can send a first request to the NEF network element, and the first request can be used to request a sample. After the NEF network element receives the first request, the NEF network element can send the first request to one or more second network elements (NWDAF). The first request includes the identification type of the sample used in the VFL task. For example, the first request can further include one or more of the following: encryption information, encrypted identification of the sample supported by the first network element (AF), or identification of the first network element (AF).

[0249] In the embodiments of the present application, the NEF network element can be transparently forwarded. It can be understood that the transparent forwarding in the embodiments of the present application can refer to forwarding without any processing, or forwarding after necessary processing (for example, format change, etc.), but the destination end can understand the valid information from the source end. In the embodiments of the present application, the service request message received by the NEF network element and the service request message sent by the NEF network element in response to the service request message can be different messages.

[0250] In an optional embodiment, the type of the sample used in the VFL task included in the first request can also be preconfigured in the first network element (NWDAF). It can be understood that if the type of the sample used in the VFL task is preconfigured / predefined / pre-set in the first network element (NWDAF), the steps S201 to S204 can not be performed.

[0251] In step S206, the second network element (NWDAF) sends a first response to the first network element (AF) through the NEF network element, and the first response includes the identification of the sample supported by the second network element. The identification of the sample supported by the second network element can be encrypted or unencrypted (such as plaintext).

[0252] Correspondingly, the first network element (AF) receives the first response.

[0253] In a possible implementation, the first request includes the type of the sample used in the VFL task. After receiving the first request, each second network element (NWDAF) can determine whether the locally stored sample identification c matches the type of the sample used in the VFL task carried in the first request. If the locally stored sample identification c of the second network element (NWDAF) does not match the type of the sample used in the VFL task, the second network element (NWDAF) can convert the sample identification c to a sample identification d that matches the type of the sample used in the VFL task. For example, assuming that the type of the sample used in the VFL task is GPSI, and the locally stored sample identification a of the second network element (NWDAF) is SUPI, the second network element (NWDAF) can request the UDM network element to complete the conversion of SUPI and GPSI. The specific conversion manner can be referred to the related description in the foregoing embodiment shown in FIG. 7, which will not be described here. If the locally stored sample identification c of the second network element (NWDAF) matches the type of the sample used in the VFL task, no conversion is needed. The second network element (NWDAF) can then send a first response to the NEF network element, and the NEF network element forwards the first response to the first network element (AF). The first response can include the unencrypted identification of the sample supported by the second network element, such as the sample identification c or the sample identification d.

[0254] In a possible implementation, the second network element (NWDAF) can also determine encryption information, and can use the encryption information to encrypt the sample identifier d to obtain an encrypted identifier Y of the sample supported by the second network element (NWDAF). Alternatively, when the sample identifier c stored locally by the second network element (NWDAF) matches the type of the identifier of the sample used in the VFL task, the second network element (NWDAF) can determine encryption information, and can use the encryption information to encrypt the sample identifier c to obtain an encrypted identifier Y of the sample supported by the second network element (NWDAF). The encryption information can include an encryption algorithm type and / or specific information of the encryption algorithm (for example, a public key, encryption, and the like). The encryption algorithm can be homomorphic encryption. The second network element (NWDAF) can further send a first response to the NEF network element, and the NEF network element forwards the first response to the first network element (AF). The first response can include the encrypted identifier Y of the sample supported by the second network element.

[0255] For example, the first response can also be referred to as a sample alignment response, a data alignment response, or the like, and the application embodiments do not limit the specific name of the first response.

[0256] In a possible implementation, the first request includes not only the identification type of the sample used in the VFL task, but also the identification X of the encrypted sample supported by the first network element (AF). After receiving the first request, each second network element (NWDAF) can determine whether the locally stored sample identification c matches the identification type of the sample used in the VFL task carried in the first request. If the sample identification c stored locally by the second network element (NWDAF) does not match the identification type of the sample used in the VFL task, the second network element (NWDAF) can convert the sample identification c into sample identification d that matches the identification type of the sample used in the VFL task. The specific conversion manner of the sample identification can be referred to the related description in the foregoing embodiment shown in FIG. 7, and will not be described here again. The second network element (NWDAF) can determine the encryption information, and can use the encryption information to perform encryption processing on the sample identification d to obtain the identification Y of the encrypted sample supported by the second network element (NWDAF). If the sample identification c stored locally by the second network element (NWDAF) matches the identification type of the sample used in the VFL task, the second network element (NWDAF) can determine the encryption information, and can use the encryption information to perform encryption processing on the sample identification c to obtain the identification Y of the encrypted sample supported by the second network element (NWDAF). The encryption information can include the encryption algorithm type and / or specific information (such as a public key, encryption) of the encryption algorithm. The encryption algorithm can be homomorphic encryption. The second network element (NWDAF) can further determine the intersection between the identification X of the encrypted sample supported by the first network element (AF) and the identification Y of the encrypted sample supported by the second network element (NWDAF) in the first request, and can send a first response to the NEF network element, and the NEF network element forwards the first response to the first network element (AF). The first response can include the intersection between the identification X of the encrypted sample supported by the first network element (AF) and the identification Y of the encrypted sample supported by the second network element (NWDAF).

[0257] In a possible implementation, the encryption information can be preconfigured, predefined, or specified by a standard, or can be determined in advance by the first network element (AF) and the second network element (NWDAF). In another possible implementation, the encryption information can also be carried in the first request. In other words, the first request can further include the encryption information. Any manner that enables the first network element (AF) and the second network element (NWDAF) to know the encryption information is within the protection scope of the present application. It can be understood that the encryption information used by the second network element (NWDAF) for encryption processing is the same as the encryption information used by the first network element (AF) for encryption processing.

[0258] It can be understood that the first network element and the second network element of the embodiments of the present application encrypt the identification of the samples supported by each other, and the identification of the samples before encryption is matched with the identification type of the samples in the VFL task. Therefore, even if the other party cannot or does not want to obtain the plaintext, the identification of the samples supported by the first network element and the second network element can be obtained by using the ciphertext after encryption. Not only the security can be improved (achieved by encryption), but also the possibility of sample identification mismatch in the sample alignment process of the vertical federated learning task can be reduced (this effect is achieved by aligning the identification type of the samples before encryption), thereby realizing the sample alignment in the vertical federated learning task.

[0259] In a possible implementation, the first request comprises not only the identification type of the sample used in the VFL task, but also the identification of the first network element (used to identify the first network element). After receiving the first request, each second network element (NWDAF) can determine whether the first request is used to request the plaintext identification of the sample or the ciphertext. If the first request is used to request the plaintext identification of the sample, the second network element (NWDAF) can determine whether to return the plaintext identification of the sample based on the type of the first network element. For example, if the first network element (i.e., the requester of the sample) is a NWDAF or a trusted AF (trusted AF), the second network element (NWDAF) can agree to return the plaintext identification of the sample. If the first network element (i.e., the requester of the sample) is an untrusted AF (untrusted AF) or an AF, the second network element (NWDAF) can not send the plaintext identification of the supported sample to the first network element, in which case the second network element (NWDAF) can refuse the request of the first network element, and of course can return the encrypted ciphertext identification, which is not limited by the embodiment of the application. If the second network element (NWDAF) determines that the plaintext identification of the sample can be returned, the second network element (NWDAF) can determine whether the sample identification c stored locally matches the identification type of the sample used in the VFL task carried in the first request. If the sample identification c stored locally by the second network element (NWDAF) does not match the identification type of the sample used in the VFL task, the second network element (NWDAF) can convert the sample identification c into sample identification d, which matches the identification type of the sample used in the VFL task. The specific conversion manner of the sample identification can be referred to the related description in the foregoing embodiment shown in FIG. 7, which is not described herein again. The second network element (NWDAF) can send a first response to the NEF network element, and the NEF network element forwards the first response to the first network element (AF), and the first response can comprise the plaintext identification of the sample supported by the second network element (NWDAF) (i.e., the sample identification d). If the sample identification c stored locally by the second network element (NWDAF) matches the identification type of the sample used in the VFL task, the second network element (NWDAF) can send a first response to the NEF network element, and the NEF network element forwards the first response to the first network element (AF), and the first response can comprise the plaintext identification of the sample supported by the second network element (NWDAF) (i.e., the sample identification c).

[0260] If the second network element (NWDAF) determines that the plaintext identification of the sample cannot be returned, the second network element (NWDAF) can refuse the first request.

[0261] If the first request is used to request the ciphertext of the sample, or the second network element (NWDAF) determines that the plaintext identifier of the sample cannot be returned, the second network element (NWDAF) can determine whether the locally stored sample identifier c matches the identifier type of the sample used in the VFL task carried in the first request. If the sample identifier c stored locally by the second network element (NWDAF) does not match the identifier type of the sample used in the VFL task, the second network element (NWDAF) can convert the sample identifier c to a sample identifier d that matches the identifier type of the sample used in the VFL task. The second network element (NWDAF) can determine the encryption information and can use the encryption information to perform encryption processing on the sample identifier d to obtain the encrypted identifier Y of the sample supported by the second network element (NWDAF). If the sample identifier c stored locally by the second network element (NWDAF) matches the identifier type of the sample used in the VFL task, the second network element (NWDAF) can determine the encryption information and can use the encryption information to perform encryption processing on the sample identifier c to obtain the encrypted identifier Y of the sample supported by the second network element (NWDAF). The second network element (NWDAF) can then send a first response to the NEF network element, and the NEF network element forwards the first response to the first network element (AF). The first response can include the encrypted identifier Y of the sample supported by the second network element, or the intersection between the encrypted identifier X of the sample supported by the first network element (AF) and the encrypted identifier Y of the sample supported by the second network element (NWDAF) (if the encrypted identifier X of the sample supported by the first network element (AF) is carried in the first request).

[0262] S207, the first network element (AF) determines the identifier of the sample commonly supported by the vertical federated learning task group based on the identifier of the sample supported by the second network element (NWDAF). The vertical federated learning task group includes the first network element (AF) and the second network element (NWDAF).

[0263] In a possible implementation manner, the implementation of step S207 in the embodiment of the present application can refer to the implementation of step S107 in the embodiment shown in FIG. 7, which will not be described here.

[0264] The AF sends a first request to the NWDAF through the NEF, and the first request includes an identification type of a sample used in a VFL task, and optionally includes an encrypted identification of a sample supported by the AF, which matches the identification type of the sample used in the VFL task; the NWDAF returns, to the AF through the NEF, a sample identification that matches the identification type of the sample used in the VFL task, and the sample identification is encrypted; or the NWDAF returns, to the AF through the NEF, an intersection between the encrypted identification of the sample supported by the AF and the identification of a sample supported by the NWDAF, and the identification of the sample supported by the NWDAF matches the identification type of the sample used in the VFL task; and the AF determines, according to the information returned by the NWDAF, an identification of a sample supported by a VFL task group (including the NWDAF and the AF). Because the identification of the sample carried in the first request by the AF and the identification of the sample returned by the NWDAF both match the identification type of the sample used in the VFL task, and both are encrypted, even if the plaintext cannot be obtained or is not obtained, the encrypted ciphertext can be used to obtain the identification of the sample supported by the NWDAF and the AF by taking the intersection. This not only improves security, but also reduces the possibility of sample identification mismatch in the sample alignment process of the vertical federated learning task, thereby achieving sample alignment in the vertical federated learning task.

[0265] Referring to FIG. 9, FIG. 9 is another flowchart of a sample alignment method in a vertical federated learning according to an embodiment of the present application. In the method, the first network element can be a NEF; the second network element can be an AF or a NWDAF, which can act as a VFL passive participant or a VFL client; and the third network element can be a NWDAF or an AF, which can act as a VFL active participant or a VFL server. The method shown in FIG. 9 takes the second network element as an AF and the third network element as a NWDAF as an example; in actual application, the second network element can be a NWDAF and the third network element can be an AF, which is not limited by the embodiments of the present application. The second network element in the method can be one or more.

[0266] As shown in FIG. 9, the sample alignment method in the vertical federated learning includes but is not limited to the following steps:

[0267] S301, the third network element (NWDAF) sends a second request to the first network element (NEF), and the second request includes one or more of the following: an identification type of a sample used in a vertical federated learning task, information used to determine the second network element (AF), or role information corresponding to the second network element.

[0268] Correspondingly, the first network element (NEF) receives the second request.

[0269] In a possible implementation, after determining the identification type of the sample used in the VFL task, the third network element (NWDAF) can send a second request to the first network element (NEF). The second request can be used to request (from the second network element) to align the sample. The second request can also be referred to as a sample alignment request, or a data alignment request, etc., and the embodiments of the present application do not limit the specific name of the second request. The second request can include but is not limited to one or more of the following: the identification type of the sample used in the VFL task, role information corresponding to the second network element (for example: VFL active participant or VFL passive participant), information used to determine the second network element (AF), or encryption information. The information used to determine the second network element can include the identification or name of the second network element, and of course can also be other information, and any information that can determine the second network element is within the protection scope of the present application. The information used to determine the second network element can be obtained through a network element discovery process. The encryption information can include the type of encryption algorithm, and / or specific information of the encryption algorithm (such as: public key, encryption), etc. The encryption algorithm can be homomorphic encryption (Homomorphic Encryption).

[0270] In a possible implementation, the second request described above can further include the identification of the sample supported by the third network element (NWDAF), and / or the identification type of the sample supported by the third network element (NWDAF). The identification of the sample supported by the third network element (NWDAF) can be encrypted or unencrypted. If the identification of the sample supported by the third network element (NWDAF) is encrypted, the third network element (NWDAF) uses the encryption information described above for encryption. For example, the identification of the sample supported by the third network element (NWDAF) can match the identification type of the sample used in the VFL task, and if the identification of the sample supported by the third network element (NWDAF) is in plaintext, it can also not match the identification type of the sample used in the VFL task.

[0271] In a possible implementation, if the second request described above includes the identification type of the sample used in the VFL task, before step S301, the sample alignment method in the vertical federated learning further includes: determining, by the third network element (NWDAF), the identification type of the sample used in the vertical federated learning task.

[0272] In a possible implementation, the identification type of the sample used in the VFL task can be pre-configured, or pre-defined, or specified by a standard protocol.

[0273] In another possible implementation, the third network element (NWDAF) and the second network element (AF) can register their capabilities to the NRF network element respectively. The third network element (NWDAF) can send a network element discovery request to the NRF network element. After receiving the network element discovery request, the NRF network element can send a network element discovery response to the third network element (NWDAF), where the network element discovery response includes information of the second network element (AF). After receiving the network element discovery response, the third network element (NWDAF) can determine the identification types of the samples used in the VFL task. The identification types of the samples used in the VFL task can be one or more of the intersection between the identification types of the samples supported by the third network element (NWDAF) and the identification types of the samples supported by the second network element (AF). For details of this implementation, refer to the related description of steps S101 to S104 in the foregoing embodiment shown in FIG. 7, which will not be repeated here.

[0274] In yet another possible implementation, the third network element (NWDAF) can determine the identification types of the samples used in the VFL task from the identification types (SUPI, GPSI, or IP address) of the samples supported by the third network element (NWDAF) itself. In other words, the identification types of the samples used in the VFL task can be a subset of the identification types of the samples supported by the third network element (NWDAF).

[0275] S302, the first network element (NEF) sends a first request to the second network element (AF), where the first request includes the identification types of the samples used in the vertical federated learning task.

[0276] Correspondingly, the second network element (AF) receives the first request.

[0277] In a possible implementation, after receiving the second request, the first network element (NEF) can send a first request to the second network element (AF). The second network element (AF) can be determined by the first network element (NEF) itself or determined according to the information for determining the second network element (AF) in the second request, which is not limited in the embodiments of the present application. The first request can be used to request samples. The first request can also be referred to as a sample acquisition request, a data acquisition request, etc., and the specific name of the first request is not limited in the embodiments of the present application. The first request can include the identification types of the samples used in the VFL task. It can be understood that the identification types of the samples used in the VFL task included in the first request can be determined by the first network element (NEF) itself or carried in the second request, which is not limited in the embodiments of the present application.

[0278] In a possible implementation, the first request can further include one or more of the following: role information (e.g., VFL primary participant or VFL passive participant) corresponding to the second network element, an analysis identifier, or encryption information. The encryption information can include an encryption algorithm type, and / or specific information of the encryption algorithm (such as a public key, encryption, etc.). The encryption algorithm can be a homomorphic encryption.

[0279] S303. The second network element (AF) sends, to the first network element (NEF), a first response including an identifier of a sample supported by the second network element. The identifier of the sample supported by the second network element matches an identifier type of the sample used in the VFL task.

[0280] Correspondingly, the first network element (NEF) receives the first response.

[0281] In a possible implementation, after receiving the first request, the second network element (AF) can determine an identifier of a sample supported by the second network element according to the identifier type of the sample used in the VFL task. For example, after receiving the first request, the second network element (AF) can determine whether a locally stored sample identifier c matches the identifier type of the sample used in the VFL task carried in the first request. If the locally stored sample identifier c of the second network element (AF) does not match the identifier type of the sample used in the VFL task, the second network element (AF) can convert the sample identifier c to a sample identifier d that matches the identifier type of the sample used in the VFL task. For example, assuming that the identifier type of the sample used in the VFL task is GPSI, and the locally stored sample identifier c of the second network element (AF) is an IP address, the second network element (AF) can map the local IP address to GPSI. The second network element (AF) can then send, to the first network element (NEF), a first response including the identifier of the sample supported by the second network element (i.e., the sample identifier d). If the locally stored sample identifier c of the second network element (AF) matches the identifier type of the sample used in the VFL task, the second network element (AF) can send, to the first network element (NEF), a first response including the identifier of the sample supported by the second network element (i.e., the sample identifier c).

[0282] In a possible implementation, the identification of the sample supported by the second network element (AF) can also be encrypted. Then, if the sample identification c stored locally by the second network element (AF) does not match the type of the identification of the sample used in the VFL task, the second network element (AF) can convert the sample identification c into a sample identification d, and can determine encryption information used to encrypt the sample identification d to obtain the encrypted identification of the sample supported by the second network element (AF). The sample identification d matches the type of the identification of the sample used in the VFL task. If the sample identification c stored locally by the second network element (AF) matches the type of the identification of the sample used in the VFL task, the second network element (AF) can determine encryption information used to encrypt the sample identification c to obtain the encrypted identification of the sample supported by the second network element (AF). The second network element (AF) can further send, to the first network element (NWDAF), a first response, where the first response can include the encrypted identification of the sample supported by the second network element.

[0283] It can be understood that if the identification of the sample supported by the second network element in the first response is encrypted, the identification of the sample supported by the third network element (NWDAF) in the second request (if any) is also encrypted, and the encryption information of both is the same.

[0284] In a possible implementation, the encryption information can be preconfigured or predefined or specified by a standard, or can be determined in advance by negotiation between the third network element (NWDAF) and the second network element (AF). In another possible implementation, the encryption information can also be carried in the first request. Any manner that enables the third network element (NWDAF) and the second network element (AF) to know the encryption information is within the protection scope of the present application.

[0285] S304, the first network element (NEF) determines, based on the identification of the sample supported by the third network element (NWDAF) and the identification of the sample supported by the second network element (AF), the identification of the sample commonly supported by the vertical federated learning task group. The vertical federated learning task group includes the third network element (NWDAF) and the second network element (AF).

[0286] In a possible implementation, after receiving the first response, the first network element (NEF) can obtain the identification (such as plaintext) of the sample supported by the third network element (NWDAF). The identification of the sample supported by the third network element (NWDAF) can be carried in the second request. If the identification of the sample supported by the third network element (NWDAF) is not carried in the second request, the first network element (NEF) can obtain the identification of the sample supported by the third network element (NWDAF) from the third network element (NWDAF).

[0287] In a possible implementation, the identifier (such as plaintext) of the sample supported by the third network element (NWDAF) obtained by the first network element (NEF) can be matched with the type of the identifier of the sample used in the VFL task, or can be not matched with the type of the identifier of the sample used in the VFL task. If the identifier of the sample supported by the third network element (NWDAF) is not matched with the type of the identifier of the sample used in the VFL task, the first network element (NEF) can obtain the type of the identifier of the sample supported by the third network element (NWDAF). The type of the identifier of the sample supported by the third network element (NWDAF) can be carried in the second request described above. According to the type of the identifier of the sample supported by the third network element (NWDAF) and the type of the identifier of the sample used in the VFL task, the first network element (NEF) can map the identifier of the sample supported by the third network element (NWDAF) to the identifier of the sample matched with the type of the identifier of the sample used in the VFL task, and then determine the identifier of the sample supported by the VFL task group. It can be understood that the identifier of the sample supported by the VFL task group is matched with the type of the identifier of the sample used in the VFL task.

[0288] In a possible implementation, the first network element (NEF) can determine the identifier of the sample supported by the VFL task group based on the identifier of the sample supported by the third network element (NWDAF) and the identifier of the sample supported by the second network element (AF). The VFL task group includes the third network element (NWDAF) and the second network element (AF). For example, the identifier of the sample supported by the VFL task group can be the intersection between the identifier (such as plaintext) of the sample supported by the third network element (NWDAF) and the identifier (such as plaintext) of the sample supported by the second network element (AF). If the second network element (AF) has M (M is an integer greater than or equal to 2) network elements, the first network element (NEF) can first determine the intersection between the identifier of the sample supported by the third network element (NWDAF) and the identifier of the sample supported by each second network element (AF) in the first response, thereby obtaining M intersections. The first network element (NEF) can then take the intersection of the M intersections as the identifier of the sample supported by the VFL task group.

[0289] In another possible implementation, if the identification of the sample supported by the second network element (AF) in the first response is encrypted, the identification of the sample supported by the third network element (NWDAF) obtained by the first network element (NEF) is also encrypted. If the second network element (AF) has M (M is an integer greater than or equal to 2) network elements, after receiving the first response, the first network element (NEF) can first determine the intersection between the identification (such as ciphertext) of the sample supported by the third network element (NWDAF) and the identification (such as ciphertext) of the sample supported by the second network element (AF) in each first response, thereby obtaining M intersections. The first network element (NEF) can then determine the identification of the sample commonly supported by the VFL task group according to the plaintext identification corresponding to the M intersections. For example, the intersection between the plaintext identification corresponding to the M intersections is taken as the identification of the sample commonly supported by the VFL task group.

[0290] In a possible implementation, after determining the identification of the sample commonly supported by the VFL task group, the first network element (NEF) can return a second response to the third network element (NWDAF), and the second response can include the result of sample alignment, such as the identification of the sample commonly supported by the VFL task group. Optionally, the first network element (NEF) can also send the result of sample alignment to the second network element (AF). Thereafter, the third network element (NWDAF) and the second network element (AF) can perform subsequent VFL training processes based on the result of sample alignment, which is described in detail in the prior art and will not be described here.

[0291] The NWDAF of the embodiment of the present application sends a second request to the NEF, which contains information for determining the AF and the identification type of the sample used in the VFL task; the NEF sends a second request to the AF based on the second request, which contains the identification type of the sample used in the VFL task; the AF returns the sample identification matching the identification type of the sample used in the VFL task to the NEF; and the NEF determines the identification of the sample commonly supported by the VFL task group (including the NWDAF and the AF) according to the sample identification returned by the AF and the sample identification supported by the NWDAF. This can reduce the possibility of sample identification mismatch in the sample alignment process of the vertical federated learning task, thereby achieving sample alignment in the vertical federated learning task.

[0292] The above describes the method of the present application in detail. In order to better implement the above-mentioned scheme of the embodiment of the present application, the embodiment of the present application also provides a corresponding device or equipment.

[0293] The embodiments of the present application can divide the functions of each network element of the present application according to the above method examples, and divide the functions of the above network elements according to the above method examples. For example, each function module can be divided according to each function, or two or more functions can be integrated in one processing module. The integrated module can be realized in the form of hardware or in the form of a software function module. It should be noted that the division of the modules in the embodiments of the present application is illustrative, and is only a logical function division. In actual implementation, there can be another division manner. The communication device of the embodiments of the present application will be described in detail below with reference to FIGS. 10 to 12.

[0294] Referring to FIG. 10, FIG. 10 is a structural schematic diagram of a communication device provided by an embodiment of the present application. As shown in FIG. 10, the communication device includes a transceiver module 10 and a processing module 20. The transceiver module 10 can realize corresponding communication functions, and the processing module 20 is used for data processing. The transceiver module 10 can also be referred to as a communication interface or a communication unit, etc.

[0295] In some embodiments of the present application, the communication device can be the first network element shown above. That is, the communication device shown in FIG. 10 can be used to perform the steps or functions performed by the first network element in the above method embodiments, etc. For example, the communication device can be a chip or a function module in the first network element, etc., and the embodiments of the present application do not limit this. The transceiver module 10 is used to perform the operations related to the transceiving of the first network element in the above method embodiments, and the processing module 20 is used to perform the operations related to the processing of the first network element in the above method embodiments.

[0296] For example, the transceiver module 10 is used to send a first request to a second network element, the first request including an identification type of a sample used in a longitudinal federated learning task; the transceiver module 10 is also used to receive a first response from the second network element, the first response including an identification of a sample supported by the second network element, the identification of the sample supported by the second network element matching the identification type of the sample used in the longitudinal federated learning task; and the processing module 20 is used to determine an identification of a sample commonly supported by a longitudinal federated learning task group based on the identification of the sample supported by the second network element, the longitudinal federated learning task group including the second network element.

[0297] It can be understood that the transceiver module 10 can send the first request to other communication devices, or the transceiver module 10 outputs the first request from the processing module 20 to other components or other function modules in the communication device, etc. The related description of the transceiver module outputting other information is similar, and will not be described in detail below.

[0298] It can be understood that the transceiver module 10 can receive the first response from other communication devices, or the transceiver module 10 inputs the first response from other components or other functional modules in the communication device, etc. The relevant description of the transceiver module inputting other information is similar, and will not be described in detail below.

[0299] Exemplarily, the transceiver module 10 is further configured to send a network element discovery request to the network storage function network element; the transceiver module 10 is further configured to receive a network element discovery response from the network storage function network element, the network element discovery response including an identification type of samples supported by a second network element; and the processing module 20 is further configured to determine an identification type of samples used in the vertical federated learning task based on the identification type of samples supported by the first network element and the identification type of samples supported by the second network element.

[0300] Exemplarily, the transceiver module 10 is further configured to send a network element discovery request to the network storage function network element, the network element discovery request including the identification type of samples supported by the first network element; the transceiver module 10 is further configured to receive a network element discovery response from the network storage function network element, the network element discovery response including information of a second network element, the identification type of samples supported by the second network element having an intersection with the identification type of samples supported by the first network element and the intersection being non-empty, the information of the second network element including the intersection. The identification type of samples used in the vertical federated learning task is the intersection.

[0301] Exemplarily, the first request further includes an encrypted identification of samples supported by the first network element; and the identification of samples supported by the second network element is an intersection between an encrypted identification of samples supported by the second network element and the encrypted identification of samples supported by the first network element.

[0302] Exemplarily, the processing module 20 is further configured to, when a first sample identification stored by the first network element does not match the identification type of samples used in the vertical federated learning task, convert the first sample identification into a second sample identification, and encrypt the second sample identification to obtain the encrypted identification of samples supported by the first network element; the second sample identification matches the identification type of samples used in the vertical federated learning task.

[0303] Exemplarily, the processing module 20 is further configured to, when a first sample identification stored by the first network element matches the identification type of samples used in the vertical federated learning task, encrypt the first sample identification to obtain the encrypted identification of samples supported by the first network element.

[0304] Exemplarily, the first request further includes encryption information. The encryption processing includes encrypting based on the encryption information.

[0305] Exemplarily, the identification of samples supported by the second network element is an encrypted identification of samples supported by the second network element.

[0306] The processing module 20 is configured to determine the identification of the samples supported by the vertical federated learning task group based on the identification of the samples supported by the first network element and the identification of the samples supported by the second network element.

[0307] The transceiver module 10 is further configured to receive a second request from a third network element, the second request including one or more of the following: the sample identification type used in the vertical federated learning task, or information used to determine the second network element. The transceiver module 10 is further configured to send the first request to the second network element according to the second request.

[0308] The second request further includes the identification of the samples supported by the third network element. The processing module 20 is configured to determine the identification of the samples supported by the vertical federated learning task group based on the identification of the samples supported by the third network element and the identification of the samples supported by the second network element, the vertical federated learning task group including the second network element and the third network element. The identification of the samples supported by the vertical federated learning task group matches the sample identification type used in the vertical federated learning task.

[0309] In the embodiments of the present application, the specific descriptions of the first request, the first response, the sample identification type used in the vertical federated learning task, the vertical federated learning task group, the identification of the samples, and the network elements can refer to the method embodiments shown in FIG. 7 or FIG. 8 or FIG. 9, which will not be repeated here.

[0310] It can be understood that the specific descriptions of the transceiver module and the processing module shown in the embodiments of the present application are only examples. For the specific functions or steps of the transceiver module and the processing module, etc., the method embodiments shown in FIG. 7 or FIG. 8 or FIG. 9 can be referred to, which will not be repeated here. In addition, the technical effects of the embodiments of the present application can refer to the technical effects of the method embodiments shown in FIG. 7 or FIG. 8 or FIG. 9, which will not be repeated here for brevity.

[0311] In other embodiments of the present application, the communication apparatus shown in FIG. 10 can be the second network element shown above. That is, the communication apparatus shown in FIG. 10 can be configured to perform the steps or functions performed by the second network element in the method embodiments shown above. The communication apparatus can be the second network element or a chip or functional module configured in the second network element, etc., which is not limited in the embodiments of the present application. The transceiver module 10 is configured to perform the operations related to the transceiving of the second network element in the method embodiments shown above, and the processing module 20 is configured to perform the operations related to the processing of the second network element in the method embodiments shown above.

[0312] Exemplarily, the transceiver module 10 is configured to receive a first request from a first network element, the first request comprising an identification type of samples used in a vertical federated learning task; and the transceiver module 10 is further configured to send a first response to the first network element, the first response comprising an identification of samples supported by a second network element, the identification of samples supported by the second network element matching the identification type of samples used in the vertical federated learning task.

[0313] Exemplarily, the processing module 20 is configured to generate various information sent by the transceiver module 10, such as the first request; and the processing module 20 is further configured to control the transceiver module 10 to send or receive various information.

[0314] Exemplarily, the first request further comprises encrypted identification of samples supported by the first network element, and the identification of samples supported by the second network element is an intersection between the encrypted identification of samples supported by the second network element and the encrypted identification of samples supported by the first network element.

[0315] Exemplarily, the processing module 20 is further configured to determine, according to the identification type of samples used in the VFL task, the encrypted identification of samples supported by the second network element that matches the identification type of samples used in the VFL task.

[0316] For example, the processing module 20 is further configured to, when the first sample identification stored by the second network element does not match the identification type of samples used in the vertical federated learning task, convert the first sample identification into a second sample identification, and encrypt the second sample identification to obtain the encrypted identification of samples supported by the second network element; the second sample identification matches the identification type of samples used in the vertical federated learning task. The processing module 20 is further configured to, when the first sample identification stored by the second network element matches the identification type of samples used in the vertical federated learning task, encrypt the first sample identification to obtain the encrypted identification of samples supported by the second network element.

[0317] Exemplarily, the first request further comprises encryption information; and the encryption processing comprises encrypting based on the encryption information.

[0318] Exemplarily, the first request further comprises an identification of the first network element. The processing module 20 is further configured to determine whether to return the plaintext identification of the samples based on a type corresponding to the identification of the first network element; and the transceiver module 10 is specifically configured to, when the second network element agrees to return the plaintext identification of the samples, send the first response to the first network element, the identification of samples supported by the second network element being the plaintext identification.

[0319] For example, the processing module 20 is further configured to determine whether the first request is used to request the plaintext identification of the samples or the ciphertext; and the processing module 20 is specifically configured to, when the first request is used to request the plaintext identification of the samples, determine whether to return the plaintext identification of the samples based on the type of the first network element.

[0320] Exemplarily, the identification of the sample supported by the second network element is an encrypted identification of the sample supported by the second network element.

[0321] Exemplarily, the identification of the sample supported by the second network element is a first sample identification stored by the second network element, and the first sample identification matches the identification type of the sample used in the vertical federated learning task.

[0322] Exemplarily, the transceiver 10 is further configured to send a network element registration request to a network storage function network element, the network element registration request including the identification type of the sample supported by the second network element; and the transceiver 10 is further configured to receive a network element registration response returned by the network storage function network element, the network element registration response being used to confirm that the network element registration is accepted.

[0323] In the embodiments of the present application, the specific descriptions of the first request, the first response, the identification type of the sample used in the vertical federated learning task, the identification of the sample, and each network element can refer to the method embodiments shown in FIG. 7 or FIG. 8 or FIG. 9, which will not be repeated here.

[0324] It can be understood that the specific descriptions of the transceiver and the processing module shown in the embodiments of the present application are only examples. For the specific functions or steps of the transceiver and the processing module, etc., it can refer to the method embodiments shown in FIG. 7 or FIG. 8 or FIG. 9, which will not be repeated here. In addition, the technical effects of the embodiments of the present application refer to the technical effects in the method embodiments shown in FIG. 7 or FIG. 8 or FIG. 9, which will not be repeated here for brevity.

[0325] Referring to FIG. 10, in some other embodiments of the present application, the communication device can be the third network element shown above. That is, the communication device shown in FIG. 10 can be used to perform the steps or functions performed by the third network element in the method embodiments above. Exemplarily, the communication device can be a third network element or a chip or a functional module configured in the third network element, etc., which is not limited in the embodiments of the present application. The transceiver 10 is used to perform the operations related to the transceiving of the third network element in the method embodiments above, and the processing module 20 is used to perform the operations related to the processing of the third network element in the method embodiments above.

[0326] Exemplarily, the processing module 20 is configured to determine the identification type of the sample used in the vertical federated learning task; and the transceiver 10 is configured to send a second request to the first network element, the second request including information used to determine the second network element and the identification type of the sample used in the vertical federated learning task.

[0327] Exemplarily, the processing module 20 is specifically configured to: control the transceiver module 10 to send a network element discovery request to the network storage function network element; control the transceiver module 10 to receive a network element discovery response from the network storage function network element, the network element discovery response including an identification type of samples supported by a second network element; and determine an identification type of samples used in the vertical federated learning task based on the identification type of samples supported by the third network element and the identification type of samples supported by the second network element.

[0328] Exemplarily, the processing module 20 is specifically configured to: control the transceiver module 10 to send a network element discovery request to the network storage function network element, the network element discovery request including an identification type of samples supported by a third network element; control the transceiver module 10 to receive a network element discovery response from the network storage function network element, the network element discovery response including information of a second network element, the second network element supporting an identification type of samples that has an intersection with the identification type of samples supported by the third network element and the intersection is not empty. The identification type of samples used in the vertical federated learning task is the intersection.

[0329] Exemplarily, the second request further includes an identification of samples supported by the third network element.

[0330] In the embodiments of the present application, the specific descriptions of the second request, the identification type of samples used in the vertical federated learning task, the network element discovery request, the network element discovery response, and each network element can be referred to the method embodiment shown in FIG. 9, and will not be repeated here.

[0331] It can be understood that the specific descriptions of the transceiver module and the processing module shown in the embodiments of the present application are only examples, and the specific functions or steps of the transceiver module and the processing module can be referred to the method embodiment shown in FIG. 9, and will not be repeated here. In addition, the technical effects of the embodiments of the present application are described in the method embodiment shown in FIG. 9, and will not be repeated here for brevity.

[0332] The communication device of the embodiments of the present application is introduced above, and possible product forms of the communication device are introduced below. It should be understood that any form of product that has the functions of the communication device described in FIG. 10 falls within the protection scope of the embodiments of the present application. It should also be understood that the following introduction is only an example, and the product form of the communication device of the embodiments of the present application is not limited to this.

[0333] In a possible implementation, in the communication apparatus shown in FIG. 10, the processing module 20 can be one or more processors, and the transceiver module 10 can be a transceiver, or the transceiver module 10 can also be a sending module and a receiving module, the sending module can be a transmitter, and the receiving module can be a receiver, and the sending module and the receiving module are integrated in one device, for example, a transceiver. In the embodiment of the present application, the processor and the transceiver can be coupled, and the connection manner of the processor and the transceiver is not limited in the embodiment of the present application. In the process of executing the above method, the process of sending information in the above method can be understood as the process of outputting the above information by the processor. When the above information is output, the processor outputs the above information to the transceiver, so that the transceiver transmits. After the above information is output by the processor, the above information can also need to be processed further, and then reaches the transceiver. Similarly, the process of receiving information in the above method can be understood as the process of receiving the input above information by the processor. When the processor receives the input information, the transceiver receives the above information and inputs the processor. Further, after the transceiver receives the above information, the above information can need to be processed further, and then inputs the processor.

[0334] Referring to FIG. 11, FIG. 11 is another structure schematic diagram of the communication apparatus provided by the embodiment of the present application. As shown in FIG. 11, the communication apparatus provided by the embodiment of the present application can be used to implement the method described in the method embodiment, and the description can be referred to the description in the method embodiment. The communication apparatus can be the first network element or the second network element or the third network element, or a chip or a circuit therein. For example, the communication apparatus includes one or more processors 1001 and a transceiver 1002. The communication apparatus can further include a memory 1003. In an implementation, the communication apparatus further includes an input and output device (not shown in the figure).

[0335] The processor 1001 is mainly used for processing communication protocol and communication data, and controlling the whole communication apparatus, executing software program, and processing data of the software program. The memory 1003 is mainly used for storing software program and data. The transceiver 1002 can include a control circuit and an antenna, and the control circuit is mainly used for converting baseband signals and radio frequency signals and processing the radio frequency signals. The antenna is mainly used for transceiving radio frequency signals in the form of electromagnetic waves. The input and output device, for example, a touch screen, a display screen, a keyboard, etc., is mainly used for receiving user input data and outputting data to the user.

[0336] When the communication apparatus is powered on, the processor 1001 can read the software program in the memory 1003, interpret and execute the instructions of the software program, and process the data of the software program. When data needs to be transmitted wirelessly, the processor 1001 performs baseband processing on the data to be transmitted, and outputs the baseband signal to the radio frequency circuit. The radio frequency circuit performs radio frequency processing on the baseband signal, and transmits the radio frequency signal in the form of electromagnetic wave through the antenna. When data is transmitted to the communication apparatus, the radio frequency circuit receives the radio frequency signal through the antenna, converts the radio frequency signal into a baseband signal, and outputs the baseband signal to the processor 1001. The processor 1001 converts the baseband signal into data and processes the data.

[0337] In another implementation, the radio frequency circuit and the antenna can be arranged independently of the processor that performs the baseband processing, for example, in a distributed scenario, the radio frequency circuit and the antenna can be arranged remotely from the communication apparatus.

[0338] The processor 1001, the transceiver 1002, and the memory 1003 can be connected through a communication bus.

[0339] For example, when the communication apparatus is configured to perform the steps or methods or functions performed by the first network element in the method embodiment shown in FIG. 7, the processor 1001 can be configured to perform step S104 and step S107 in FIG. 7, and / or other processes of the technology described herein; the transceiver 1002 can be configured to perform step S101, step S102, and step S105 in FIG. 7, and / or other processes of the technology described herein.

[0340] For example, when the communication apparatus is configured to perform the steps or methods or functions performed by the second network element in the method embodiment shown in FIG. 7, the processor 1001 can be configured to generate the first response, and / or other processes of the technology described herein; the transceiver 1002 can be configured to perform step S106 in FIG. 7, and / or other processes of the technology described herein.

[0341] For example, when the communication apparatus is configured to perform the steps or methods or functions performed by the NRF network element in the method embodiment shown in FIG. 7, the processor 1001 can be configured to perform step S101 in FIG. 7, and / or other processes of the technology described herein; the transceiver 1002 can be configured to perform step S103 in FIG. 7, and / or other processes of the technology described herein.

[0342] For example, when the communication apparatus is configured to perform the steps or methods or functions performed by the first network element in the method embodiment shown in FIG. 8, the processor 1001 can be configured to perform step S204 and step S207 in FIG. 8, and / or other processes described herein; the transceiver 1002 can be configured to perform step S201, step S202 and step S205 in FIG. 8, and / or other processes described herein.

[0343] For example, when the communication apparatus is configured to perform the steps or methods or functions performed by the second network element in the method embodiment shown in FIG. 8, the processor 1001 can be configured to generate the first response, and / or other processes described herein; the transceiver 1002 can be configured to perform step S201 and step S206 in FIG. 8, and / or other processes described herein.

[0344] For example, when the communication apparatus is configured to perform the steps or methods or functions performed by the NRF network element in the method embodiment shown in FIG. 8, the processor 1001 can be configured to perform step S201 in FIG. 8, and / or other processes described herein; the transceiver 1002 can be configured to perform step S203 in FIG. 8, and / or other processes described herein.

[0345] For example, when the communication apparatus is configured to perform the steps or methods or functions performed by the first network element in the method embodiment shown in FIG. 9, the processor 1001 can be configured to perform step S304 in FIG. 9, and / or other processes described herein; the transceiver 1002 can be configured to perform step S302 in FIG. 9, and / or other processes described herein.

[0346] For example, when the communication apparatus is configured to perform the steps or methods or functions performed by the second network element in the method embodiment shown in FIG. 9, the processor 1001 can be configured to generate the first response, and / or other processes described herein; the transceiver 1002 can be configured to perform step S303 in FIG. 9, and / or other processes described herein.

[0347] For example, when the communication apparatus is configured to perform the steps or methods or functions performed by the third network element in the method embodiment shown in FIG. 9, the processor 1001 can be configured to generate the first request, and / or other processes described herein; the transceiver 1002 can be configured to perform step S301 in FIG. 9, and / or other processes described herein.

[0348] In any of the implementation manners above, the processor 1001 can include a transceiver for implementing the receiving and sending functions. For example, the transceiver can be a transceiver circuit, or an interface, or an interface circuit. The transceiver circuit, interface, or interface circuit for implementing the receiving and sending functions can be separate or integrated together. The transceiver circuit, interface, or interface circuit above can be used for reading and writing of codes / data, or the transceiver circuit, interface, or interface circuit above can be used for transmission or transfer of signals.

[0349] In any of the implementation manners above, the processor 1001 can store instructions, which can be a computer program, and the computer program can run on the processor 1001 to enable the communication apparatus to perform the methods described in the method embodiments above. The computer program can be fixed in the processor 1001, and in this case, the processor 1001 can be implemented by hardware.

[0350] In an implementation manner, the communication apparatus can include a circuit, which can implement the functions of sending or receiving or communicating in the foregoing method embodiments. The processor and the transceiver described in the present application can be implemented on an integrated circuit (IC), an analog IC, a radio frequency integrated circuit (RFIC), a mixed-signal IC, an application specific integrated circuit (ASIC), a printed circuit board (PCB), an electronic device, etc. The processor and the transceiver can also be manufactured by various IC technologies, such as complementary metal oxide semiconductor (CMOS), n metal-oxide-semiconductor (NMOS), positive channel metal oxide semiconductor (PMOS), bipolar junction transistor (BJT), bipolar CMOS (BiCMOS), silicon germanium (SiGe), gallium arsenide (GaAs), etc.

[0351] It can be understood that the communication apparatus shown in the embodiments of the present application can also have more components than those shown in FIG. 11, and the embodiments of the present application do not limit this. The methods performed by the processor and the transceiver shown above are only examples, and for the specific steps performed by the processor and the transceiver, reference can be made to the description of the method embodiments above.

[0352] In another possible implementation, in the communication apparatus shown in FIG. 10, the processing module 20 can be one or more logic circuits, and the transceiver module 10 can be an input / output interface, also referred to as a communication interface, or an interface circuit, or an interface, etc. Alternatively, the transceiver module 10 can also be a sending module and a receiving module, the sending module can be an output interface, and the receiving module can be an input interface, and the sending module and the receiving module are integrated in one module, for example, an input / output interface. Referring to FIG. 12, FIG. 12 is another structural schematic diagram of a communication apparatus provided by an embodiment of the present application. As shown in FIG. 12, the communication apparatus can be any one of the aforementioned first network element, second network element, or third network element. The communication apparatus shown in FIG. 12 includes a processor 901 and an interface circuit 902. That is, the aforementioned processing module 20 can be implemented by the processor 901, and the transceiver module 10 can be implemented by the interface circuit 902. Among them, the processor 901 can be a chip, a processing circuit, an integrated circuit, or a system on chip (SoC) chip, etc., and the interface circuit 902 can be a communication interface circuit, an input / output interface circuit, a pin, etc. For example, FIG. 12 is an example in which the aforementioned communication apparatus is a chip, and the chip includes the processor 901 and the interface circuit 902.

[0353] In the embodiments of the present application, the processor and the interface circuit can also be coupled to each other. The specific connection mode of the processor and the interface circuit is not limited in the embodiments of the present application.

[0354] For example, when the communication apparatus is used to execute the method or function or step executed by the first network element in any of the preceding method embodiments, the interface circuit 902 is configured to output the first request; the interface circuit 902 is configured to input the first response; and the processor 901 is configured to determine the identification of the samples commonly supported by the vertical federated learning task group based on the identification of the samples supported by the second network element.

[0355] For example, when the communication apparatus is used to execute the method or function or step executed by the second network element in any of the preceding method embodiments, the interface circuit 902 is configured to input the first request; and the interface circuit 902 is configured to output the first response.

[0356] For example, when the communication apparatus is used to execute the method or function or step executed by the third network element in the method embodiment shown in FIG. 9, the processor 901 is configured to determine the identification type of the samples used in the vertical federated learning task; and the interface circuit 902 is configured to output the second request.

[0357] In the embodiments of the present application, the specific description of the first request, the first response, the identification of the samples, the identification of the samples commonly supported by the vertical federated learning task group, the identification type of the samples used in the vertical federated learning task, the second request, etc. can refer to the method embodiments shown above, which will not be repeated here.

[0358] It can be understood that the communication apparatuses shown in the embodiments of the present application can implement the methods provided by the embodiments of the present application in the form of hardware, or implement the methods provided by the embodiments of the present application in the form of software, and the like, which are not limited in the embodiments of the present application.

[0359] For the specific implementation of each embodiment shown in FIG. 12, reference can be made to the above-mentioned embodiments, which will not be described in detail here.

[0360] The embodiments of the present application further provide a communication system, which comprises a first network element and a second network element, and optionally comprises a third network element, the first network element and the second network element can be used to execute the method in the method embodiments shown in FIG. 7 or FIG. 8, and the first network element, the second network element, and the third network element can be used to execute the method in the method embodiment shown in FIG. 9.

[0361] In addition, the present application further provides a computer program, which is used to implement the operations and / or processes executed by the above-mentioned network elements in the methods provided by the present application.

[0362] The present application further provides a computer readable storage medium, which stores computer codes, when the computer codes are run on a computer, the computer codes make the computer execute the operations and / or processes executed by the above-mentioned network elements in the methods provided by the present application.

[0363] The present application further provides a computer program product, which comprises computer codes or computer programs, when the computer codes or computer programs are run on a computer, the operations and / or processes executed by the above-mentioned network elements in the methods provided by the present application are executed.

[0364] In several embodiments provided by the present application, it should be understood that the disclosed system, device and method can be implemented by other means. For example, the above-mentioned device embodiments are only illustrative, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interfaces, devices or units, and can also be electrical, mechanical or other form of connection.

[0365] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the technical effects of the scheme provided by the embodiments of the present application.

[0366] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present alone, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0367] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part of the prior art that contributes to the technical solutions, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a readable storage medium, including a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned readable storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0368] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A sample alignment method in federated learning, comprising: The method comprises: The first network element sends a first request to a second network element, the first request comprising an identification type of samples used in a vertical federated learning task; The first network element receives a first response from the second network element, the first response comprising an identification of samples supported by the second network element, the identification of samples supported by the second network element matching the identification type of samples used in the vertical federated learning task; The first network element determines an identification of samples commonly supported by a vertical federated learning task group based on the identification of samples supported by the second network element, the vertical federated learning task group comprising the second network element.

2. The method of claim 1, wherein, Before the first network element sends the first request to the second network element, the method further comprises: The first network element sends a network element discovery request to a network storage function network element; The first network element receives a network element discovery response from the network storage function network element, the network element discovery response comprising an identification type of samples supported by a second network element; The first network element determines the identification type of samples used in the vertical federated learning task based on the identification type of samples supported by the first network element and the identification type of samples supported by the second network element.

3. The method according to claim 1 or 2, characterized in that, Before the first network element sends the first request to the second network element, the method further comprises: The first network element sends a network element discovery request to a network storage function network element, the network element discovery request comprising an identification type of samples supported by the first network element; The first network element receives a network element discovery response from the network storage function network element, the network element discovery response comprising information of a second network element, the identification type of samples supported by the second network element having a non-empty intersection with the identification type of samples supported by the first network element; the information of the second network element comprising the intersection; The identification type of samples used in the vertical federated learning task is the intersection.

4. The method according to any one of claims 1 to 3, characterized in that, The first request further comprises an encrypted identification of samples supported by the first network element; the identification of samples supported by the second network element is an intersection between an encrypted identification of samples supported by the second network element and the encrypted identification of samples supported by the first network element.

5. The method of claim 4, wherein, Before the first network element sends the first request to the second network element, the method further comprises: When a first sample identification stored by the first network element does not match the identification type of samples used in the vertical federated learning task, the first network element converts the first sample identification into a second sample identification and performs encryption processing on the second sample identification to obtain an encrypted identification of samples supported by the first network element; The second sample identification matches the identification type of samples used in the vertical federated learning task.

6. The method of claim 5, wherein, The method further comprises: When a first sample identification stored by the first network element matches the identification type of samples used in the vertical federated learning task, the first network element performs encryption processing on the first sample identification to obtain an encrypted identification of samples supported by the first network element.

7. The method according to claim 5 or 6, characterized in that, The first request further comprises encryption information; The encryption processing comprises encryption based on the encryption information.

8. The method according to any one of claims 1 to 3, characterized in that, The identification of samples supported by the second network element is an encrypted identification of samples supported by the second network element.

9. The method of claim 8, wherein, The first network element determines, based on the identification of the samples supported by the second network element, the identification of the samples commonly supported by the vertical federated learning task group, including: The first network element determines, based on the identification of the samples supported by the second network element, the identification of the samples commonly supported by the vertical federated learning task group, including:

10. The method of claim 1, wherein, Before the first network element sends the first request to the second network element, the method further includes: The first network element receives a second request from a third network element, and the second request includes one or more of the following: a sample identification type used in the vertical federated learning task, or information used to determine the second network element.

11. The method of claim 10, wherein, The second request further includes the identification of the samples supported by the third network element; The first network element determines, based on the identification of the samples supported by the second network element, the identification of the samples commonly supported by the vertical federated learning task group, including: The first network element determines, based on the identification of the samples supported by the third network element and the identification of the samples supported by the second network element, the identification of the samples commonly supported by the vertical federated learning task group, and the identification of the samples commonly supported by the VFL task group matches the sample identification type used in the VFL task, and the vertical federated learning task group includes the second network element and the third network element.

12. The method of claim 2, wherein, Before the first network element sends a network element discovery request to a network storage function network element, the method further includes: The second network element sends a network element registration request to a network storage function network element, and the network element registration request includes an identification type of the samples supported by the second network element; The second network element receives a network element registration response returned by the network storage function network element, and the network element registration response is used to confirm that network element registration is accepted.

13. A sample alignment method in federated learning, comprising: including: The second network element receives a first request from a first network element, and the first request includes an identification type of samples used in a vertical federated learning task; The second network element sends a first response to the first network element, and the first response includes the identification of the samples supported by the second network element, which matches the identification type of the samples used in the vertical federated learning task.

14. The method of claim 13, wherein, The first request further includes encrypted identification of the samples supported by the first network element, and the identification of the samples supported by the second network element is an intersection between encrypted identification of the samples supported by the second network element and encrypted identification of the samples supported by the first network element.

15. The method of claim 14, wherein, Before the second network element sends the first response to the first network element, the method further includes: The second network element determines, according to the identification type of the samples used in the vertical federated learning task, the identification of the samples supported by the second network element that matches the identification type of the samples used in the vertical federated learning task.

16. The method of claim 15, wherein, The second network element determines, according to the identification type of the samples used in the vertical federated learning task, the identification of the samples supported by the second network element that matches the identification type of the samples used in the vertical federated learning task, including: When the first sample identifier stored by the second network element does not match the type of sample identifier used in the federated learning task, the second network element converts the first sample identifier into a second sample identifier, and encrypts the second sample identifier to obtain an encrypted sample identifier supported by the second network element. The second sample identifier matches the type of sample identifier used in the federated learning task.

17. The method of claim 16, wherein, The method further includes: When the first sample identifier stored by the second network element matches the type of sample identifier used in the federated learning task, the second network element encrypts the first sample identifier to obtain an encrypted sample identifier supported by the second network element.

18. The method of claim 16 or 17, wherein, The first request further includes encryption information. The encryption processing includes encrypting based on the encryption information.

19. The method of claim 14, wherein, The first request further includes an identifier of the first network element. Before the second network element sends a first response to the first network element, the method further includes: The second network element determines, based on the type of the first network element, a plaintext identifier of the returned sample. The sample identifier supported by the second network element is a plaintext identifier.

20. The method of claim 19, wherein, Before the second network element determines, based on the type of the first network element, a plaintext identifier of the returned sample, the method further includes: The second network element determines that the first request is used to request a plaintext identifier of a sample.

21. The method of claim 13, wherein, The sample identifier supported by the second network element is an encrypted sample identifier supported by the second network element.

22. The method of claim 13, wherein, The sample identifier supported by the second network element is the first sample identifier stored by the second network element, which matches the type of sample identifier used in the federated learning task.

23. A sample alignment method in federated learning, comprising: includes: The third network element determines the type of sample identifier used in the federated learning task. The third network element sends a second request to the first network element, the second request including one or more of the following: the type of sample identifier used in the federated learning task, or information used to determine the second network element.

24. The method of claim 23, wherein, The third network element determines the type of sample identifier used in the federated learning task, including: The third network element sends a network element discovery request to a network storage function network element; The third network element receives a network element discovery response from the network storage function network element, the network element discovery response including the type of sample identifier supported by the second network element; The third network element determines the type of sample identifier used in the federated learning task based on the type of sample identifier supported by the third network element and the type of sample identifier supported by the second network element.

25. The method of claim 23, wherein, The third network element determines the type of sample identifier used in the federated learning task, including: The third network element sends a network element discovery request to a network storage function network element, the network element discovery request including the type of sample identifier supported by the third network element; The third network element receives a network element discovery response from the network storage function network element, the network element discovery response including information of the second network element, the type of sample identifier supported by the second network element having a non-empty intersection with the type of sample identifier supported by the third network element; The type of sample identifier used in the federated learning task is the intersection.

26. The method of any one of claims 23-25, wherein, The second request further comprises an identification of samples supported by the third network element.

27. A communications device, characterized by comprising means for performing the method of any of claims 1 to 26.

28. A readable storage medium, characterized by, The readable storage medium has stored therein program instructions, which when executed on a communication device, cause the communication device to perform the method of any of claims 1 to 26.

29. A computer program product, characterised in that, The computer program product, when executed, performs the method of any of claims 1 to 26.

30. A communication system, characterized by comprising one or more of the following network elements: a first network element for performing the method of any of claims 1 to 12, or a second network element for performing the method of any of claims 13 to 21; Alternatively, the communication system comprises one or more of the following network elements: a first network element for performing the method of any of claims 1 to 12, a second network element for performing the method of any of claims 13 to 21, or a third network element for performing the method of any of claims 22 to 25.

Images