Search

CN-121997997-A - Communication method, device and readable storage medium in longitudinal federal learning

CN121997997ACN 121997997 ACN121997997 ACN 121997997ACN-121997997-A

Abstract

The application relates to the technical field of communication, in particular to a communication method, a device and a readable storage medium in longitudinal federal learning, wherein the method comprises the steps that an AF sends a first request message to a NEF, wherein the first request message comprises a first identifier of a first NWDAF and an identifier of a VFL task, the NEF sends a second request message to a UDR, wherein the identifier of the VFL task is included, the UDR returns a second response message to the NEF, wherein the mapping relation between the first identifier of each NWDAF in one or more NWDAF and the second identifier of each MWDAF is included, the one or more NWDAF participates in the VFL task, the one or more NWDAF comprises a first NWDAF, and the NEF sends a third request message to the first NWDAF, wherein the third request message comprises a second identifier of the first NWDAF. By adopting the application, the service provided by NWDAF can be obtained, and the safety of NWDAF can be improved.

Inventors

  • GUO TAO
  • WU YIZHUANG
  • GUO YANFEI
  • LEI ZHONGDING
  • HUANG LI

Assignees

  • 华为技术有限公司

Dates

Publication Date
20260508
Application Date
20241108

Claims (20)

  1. 1. A method of communication in vertical federal learning, comprising: Receiving a first request message, the first request message including a first identification of a first network data analysis function NWDAF and an identification of a Vertical Federal Learning (VFL) task, the VFL task being for training a model, the first request message being for requesting a service provided by the model; sending a second request message to a data processing network element, the second request message including an identification of the VFL task; Receiving a second response message from the data processing network element, the second response message comprising a mapping of a first identity of each NWDAF of one or more NWDAF to a second identity of each NWDAF, the one or more NWDAF engaged in the VFL task, the one or more NWDAF comprising the first NWDAF; and sending a third request message to the first NWDAF according to the second identifier of the first NWDAF, where the third request message is used to request a service provided by the model.
  2. 2. The method of claim 1, wherein the first identity is used for identity NWDAF outside a public land mobile network, PLMN, and the second identity is used for identity NWDAF inside the PLMN.
  3. 3. The method according to claim 1 or 2, wherein before said sending the second request message to the data processing network element, the method further comprises: And determining that the second identifier corresponding to the first identifier of the first NWDAF is not locally available.
  4. 4. A method according to any of claims 1 to 3, wherein after said receiving a second response message from the data processing network element, the method further comprises: storing a mapping of the first identity of each NWDAF and the second identity of each NWDAF of the one or more NWDAF.
  5. 5. The method of claim 4, wherein after receiving the second response message from the data processing network element, the method further comprises: receiving a fourth request message, the fourth request message including a first identification of a second NWDAF; If the one or more NWDAF includes the second NWDAF, a fifth request message is sent to the second NWDAF, according to a second identification of the second NWDAF, the fifth request message being for requesting services provided by the model.
  6. 6. The method according to any one of claims 1 to 5, wherein prior to the receiving the first request message, the method further comprises: Receiving a first message comprising a first identification of each NWDAF of the one or more NWDAF, and an identification of the VFL task; obtaining a second identification of participation in the one or more NWDAF; a storage request is sent to the data processing network element, the storage request including a first identification of the one or more NWDAF, an identification of the VFL task, and a second identification of the one or more NWDAF.
  7. 7. The method of claim 6, wherein prior to the receiving the first message, the method further comprises: receiving a network element discovery request, wherein the network element discovery request is used for discovering candidates NWDAF, and the network element discovery request comprises an analysis identifier; Assigning a first identification to the candidate NWDAF; Sending a network element discovery response to the application function AF, wherein the network element discovery response comprises a first identification of the candidate NWDAF; Wherein the candidate NWDAF includes the one or more NWDAF, the VFL task is to train a model to which the analysis identifier corresponds.
  8. 8. A method of communication in vertical federal learning, comprising: Receiving a second request message, the second request message including an identification of a Vertical Federal Learning (VFL) task, the VFL task being for training a model; Sending a second response message to the first network opening function NEF, the second response message comprising a mapping of the first identity of each NWDAF and the second identity of each NWDAF of one or more network data analysis functions NWDAF, the one or more NWDAF participating in the VFL task, the one or more NWDAF comprising the first NWDAF.
  9. 9. The method of claim 8, wherein the first identity is used for identity NWDAF outside a public land mobile network, PLMN, and the second identity is used for identity NWDAF inside the PLMN.
  10. 10. The method according to claim 8 or 9, wherein prior to the receiving the second request message, the method further comprises: Receiving a storage request from the first NEF, the storage request including a first identification and a second identification of each NWDAF of the one or more NWDAF, and an identification of the VFL task; a mapping of the first and second identifications participating in each NWDAF of the one or more NWDAF and the identification of the VFL task is stored.
  11. 11. The method according to any one of claims 9 to 10, further comprising: Receiving a fifth request message from the second NEF, the fifth request message including an identification of the VFL task; A fifth response message is sent to the second NEF, the fifth response message including a mapping of the first identity of each NWDAF of the one or more NWDAF to the second identity of each NWDAF, the one or more NWDAF to participate in the VFL task, the one or more NWDAF including the first NWDAF.
  12. 12. The method of claim 10 or 11, wherein the one or more NWDAF further comprise a second NWDAF.
  13. 13. A method of communication in vertical federal learning, comprising: receiving a network element discovery request, wherein the network element discovery request comprises information of an application function AF, and the network element discovery request is used for discovering a candidate network data analysis function NWDAF; Determining a candidate NWDAF according to the network element discovery request, wherein the candidate NWDAF can provide service for the AF; and sending a network element discovery response to the network opening function NEF, wherein the network element discovery response comprises the candidate NWDAF information.
  14. 14. The method of claim 13, wherein the network element discovery request further comprises one or more of an analysis identity, or a longitudinal federal learning VFL capability type, wherein the candidate NWDAF matches the VFL capability type, and wherein the analysis identity supported by the candidate NWDAF is the same as the analysis identity in the network element discovery request.
  15. 15. The method according to claim 13 or 14, wherein the information of the AF is a type or an identification or a domain name of the AF; the candidate NWDAF allowed network function consumer list includes an identification of the AF, or the candidate NWDAF allowed network function consumer domain name includes a domain name of the AF, or the candidate NWDAF allowed network function consumer type includes a type of the AF.
  16. 16. The method according to any of claims 13 to 15, wherein the network element discovery request further comprises indication information indicating that a first identity is allocated; After receiving the network element discovery request, the method further includes: and determining to allocate a first identifier for the candidate NWDAF according to the indication information, wherein the information of the candidate NWDAF comprises the first identifier and the second identifier of the candidate NWDAF.
  17. 17. The method of claim 16, wherein the first identity is used for identity NWDAF outside a public land mobile network, PLMN, and the second identity is used for identity NWDAF inside the PLMN.
  18. 18. A method of communication in vertical federal learning, comprising: The application function AF sends a longitudinal federal learning VFL training request to the network open function NEF, the VFL training request including a first identification of the network data analysis function NWDAF participating in a VFL task for training a model, the VFL training request for requesting NWDAF of the participating VFL task to train the model; The AF receives the VFL training response from the NEF; the AF stores the mapping relation between the identification of the NEF and the identification of the VFL task; the AF sends a VFL reasoning request to the NEF corresponding to the identification of the VFL task, wherein the VFL reasoning request comprises the first identification, and the VFL reasoning request is used for requesting the service provided by the model.
  19. 19. A communication device comprising means for performing the method of any one of claims 1 to 18.
  20. 20. A readable storage medium, characterized in that the readable storage medium has stored therein program instructions, which when run on a communication device, cause the communication device to perform the method of any of claims 1 to 18.

Description

Communication method, device and readable storage medium in longitudinal federal learning Technical Field The present application relates to the field of communications technologies, and in particular, to a method and an apparatus for communication in longitudinal federal learning, and a readable storage medium. Background Federal learning (FEDERATED LEARNING, FL) can be categorized into lateral federal learning (horizontal FEDERATED LEARNING), longitudinal federal learning (VERTICAL FEDERATED LEARNING), and federal migration learning (FEDERATED TRANSFER LEARNING). Wherein longitudinal federal learning (VERTICAL FEDERATED LEARNING, VFL) is generally applicable to scenarios where training samples of participating entities (e.g., network elements in the longitudinal federation) overlap more and data features overlap less. Currently, the third generation partnership project (3rd generation partnership project,3GPP) supports training and/or reasoning through longitudinal federal learning to implement mobile network-related analytics services. The application function (application function, AF) may act as a VFL server (VFL SERVER), while the AF may or may not be trusted. How to securely request a network DATA ANALYTICS function (NWDAF) as a VFL client (VFL CLIENT) to provide services when an AF as a VFL server is not trusted is a matter of ongoing investigation by those skilled in the art. Disclosure of Invention The embodiment of the application provides a communication method, a device and a readable storage medium in longitudinal federal learning, which can not only obtain services provided by a VFL client (such as NWDAF) but also improve the security of the VFL client under the condition that a VFL server (such as AF) is not trusted. The application is described below in terms of various aspects, with the understanding that the embodiments and advantages of the various aspects below may be referenced to one another. In a first aspect, the present application provides a method of communication in vertical federal learning, which method may be performed by a communication device, which may be a network open function (network exposure function, NEF) or a chip in the NEF. For ease of description, the communication device is herein described as a NEF. The method comprises the steps that the NEF receives a first request message from the AF, wherein the first request message comprises a first identifier of a first NWDAF and an identifier of a VFL task, sends a second request message to the data processing network element, the second request message comprises the identifier of the VFL task, receives a second response message from the data processing network element, the second response message comprises a mapping relation between the first identifier of each NWDAF in one or more NWDAF and the second identifier of each NWDAF, the one or more NWDAF participates in the VFL task, the one or more NWDAF comprises a first NWDAF, and sends a third request message to the first NWDAF according to the mapping relation. Wherein the VFL task is used to train the model. The first request message is for requesting a service provided by the VFL task trained model. Or the first request message is used to request the model corresponding to the identification of the VFL task to provide service. The first request message may be a VFL prediction/reasoning request, or a sample alignment request, for example. The second request message may be used to obtain a first identification and a second identification of NWDAF participating in the VFL task. The third request message is also used for requesting the service provided by the model, or the third request message is used for requesting the service provided by the model corresponding to the identification of the VFL task. Illustratively, the third request message includes a second identifier corresponding to the first identifier of the first NWDAF. In the present application, the first identifier may be an identifier for identifying NWDAF outside the public land mobile network (public land mobile network, PLMN), or the first identifier may be an identifier for identifying NWDAF outside the public land mobile network (public land mobile network, PLMN), or the first identifier may be an identifier for identifying NWDAF an AF outside the PLMN, or the first identifier may be an identifier for identifying NWDAF an untrusted AF. For example, the first identification may be a temporary NF ID (temporal NF INSTANCE ID). For example, the same NWDAF may have multiple first identities that may be used by different devices that are open to the outside of the PLMN, such as the multiple first identities may be multiple different AF identities of the NWDAF. The plurality of first identifications may also be identifying the NWDAF for the same AF that uses different first identifications in different services or tasks (e.g., VFL tasks). Correspondingly, the second identifier in the present application may be