Network node of mobile communication network, server, and computer-readable storage medium

Abstract

This server comprises: a transmission means configured to transmit, to a second network node of a mobile communication network, a first message indicating participation in vertical federated learning to be performed with a first network node of the mobile communication network; and a generation means configured to generate a learning model in the vertical federated learning. The first message includes propriety information indicating whether or not the learning model generated in the vertical federated learning can be transmitted to the mobile communication network.

Description

Network node, server and computer-readable storage medium for mobile communication network

[0001] The present disclosure relates to a technology for utilizing learning models in a mobile communication system.

[0002] Non-Patent Document 1 discloses a configuration for a data analysis service in a mobile communication network. In the following description, a network function (NF) that provides a service is referred to as a producer NF, and an NF that uses (subscribes to) a service provided by the producer NF is referred to as a consumer NF. Note that a producer NF can also operate as a consumer NF at the same time.

[0003] According to Non-Patent Document 1, a network data analysis function (NWDAF) that is a producer NF that provides a data analysis service is provided in a core network of a mobile communication network. A consumer NF obtains analysis results by subscribing to the data analysis service provided by the NWDAF.

[0004] The consumer NF is implemented, for example, in an external data network (DN) connected to the mobile communication network, for example, in a server located on the Internet. The server is operated, for example, by a service provider that provides a specific service, for example, a video distribution service, to wireless devices (WDs) on the mobile communication network. The service provider can obtain analysis results, such as the quality of experience (QoE), of users of WDs that use the service provider's services from the mobile communication network by subscribing the server (NF) operated by the service provider to a data analysis service.

[0005] In order to obtain the analysis results of the quality of experience, etc., information held by the mobile communication network (hereinafter referred to as NW information) and information held by the service provider, such as information on the application for using the service (hereinafter referred to as APP information), can be used. When using both NW information and APP information for analysis, the NWDAF performs inference using the NW information collected within the mobile communication network and the APP information notified from the NF as inputs to a learning model, and notifies the NF of the result (analysis result).

[0006] However, since the APP information may include privacy information that the service provider cannot submit to the outside, there may be cases where the APP information cannot be transmitted to the NWDAF. Therefore, it is also possible to consider a configuration in which the NWDAF distributes a learning model to the NF, and the NF performs inference using the learning model. However, since the NW information may also include privacy information that the mobile communication network operator cannot submit to the outside, there may be cases where the NW information cannot be transmitted to the NF.

[0007] In such a case, the vertical federated learning (VFL) configuration disclosed in Non-Patent Document 2 can be applied. Specifically, as shown in FIG. 1 , the NWDAF generates intermediate information using network information as an input to a first learning model and transmits the intermediate information to the NF. The NF performs inference using the intermediate information acquired from the NWDAF and APP information collected by the NF as input to a second learning model. By applying the VFL configuration, the information transmitted from the NWDAF to the NF can be treated as intermediate information different from the network information, and the NF can perform inference.

[0008] 3GPP TS23.288, V18.3.0, September 2023 3GPP S2-2306424

[0009] In the configuration of Fig. 1, in order to learn (update) the first learning model and the second learning model, the NF determines the error in the output of the second learning model to update the second learning model, and also calculates the error in the intermediate information and feeds this error in the intermediate information back to the NWDAF, which then updates the first learning model based on the error in the intermediate information.

[0010] For example, the second learning model generated (trained) by the NF in the configuration shown in Figure 1 can be reused for analysis in an NF operated by a service provider different from the service provider operating the NF. However, a procedure for reusing a second learning model generated by one NF in another NF has not been defined.

[0011] According to one aspect of the present disclosure, a server comprises a transmitting means configured to transmit a first message to a second network node of a mobile communication network indicating participation in vertical federated learning with a first network node of the mobile communication network, and a generating means configured to generate a learning model in the vertical federated learning, wherein the first message includes availability information indicating whether the learning model generated in the vertical federated learning can be transmitted to the mobile communication network.

[0012] Other features and advantages of the present invention will become apparent from the following description taken in conjunction with the accompanying drawings, in which the same or similar elements are designated by the same reference numerals.

[0013] FIG. 1 is an explanatory diagram of VFL, a system configuration diagram, a diagram showing an example of a sequence, a diagram showing an example of a sequence, a diagram showing an example of a configuration of a network node, and a diagram showing an example of a configuration of a server.

[0014] Hereinafter, the embodiments will be described in detail with reference to the accompanying drawings. Note that the following embodiments do not limit the scope of the invention as claimed, and not all combinations of features described in the embodiments are necessarily essential to the invention. Two or more of the features described in the embodiments may be arbitrarily combined. Furthermore, the same reference numerals are used for the same or similar components, and redundant explanations will be omitted.

[0015] Fig. 2 is a configuration diagram of a communication system according to this embodiment. According to Fig. 2, a core network 100 of a mobile communication network includes a network node 1 that implements an NWDAF and a network node 2 that implements a network repository function (NRF). Note that the network node 1 and the network node 2 may be the same network node. In the following description, the network node 1 is also referred to as an NWDAF, and the network node 2 is also referred to as an NRF.

[0016] The NWDAF can operate as a producer NF that provides analysis services to consumer NFs inside the mobile communication network and consumer NFs outside the mobile communication network. Specifically, the NWDAF can perform inference using a learning model and provide inference results (analysis results) to the consumer NFs. Furthermore, the NWDAF can provide intermediate information to the consumer NFs instead of inference results, as shown in FIG. 1.

[0017] The NRF has a function of intermediating between producer NFs in a mobile communication network and consumer NFs that want to use services provided by the producer NFs. Note that the core network 100 also includes various network functions of types different from the NWDAF and NRF.

[0018] 2 are located in a DN connected to the core network 100, for example, the Internet, and are operated by a service provider that provides services such as video distribution services to WDs of a mobile communication network. Alternatively, the servers 3 and 4 are located in a mobile communication network other than the mobile communication network that operates the core network 100, and are operated by an operator of the other mobile communication network. The servers 3 and 4 implement consumer NFs that can use the analysis services provided by the NWDAF. In the following description, the NF implemented in the server 3 will also be referred to as a first NF, and the NF implemented in the server 4 will also be referred to as a second NF.

[0019] In this embodiment, the first NF is an NF capable of generating and learning a second learning model in the VFL configuration shown in FIG. 1 , and the second NF is an NF that receives and uses the second learning model in the VFL configuration shown in FIG. 1 . For example, the second NF acquires the second learning model generated by the first NF and performs analysis using the acquired second learning model. Note that whether a consumer NF that uses the analysis service provided by the NWDAF is configured as the first NF or the second NF can be determined by the operator operating the consumer NF. Furthermore, depending on the analysis content, one consumer NF can be either the first NF or the second NF. For example, one consumer NF can use the second learning model generated by the VFL with the NWDAF for analyzing quality of experience, and use the second learning model generated by another NF for other analyses.

[0020] 3 is a flowchart of a process in which the NWDAF and the first NF perform VFL to generate a first learning model and a second learning model. In S10, the first NF sends a registration request message indicating participation in VFL to the NRF. The registration request message includes reuse information. The reuse information may include descriptive information that identifies the analysis content, network information used as input for the first learning model, and application information used as input for the second learning model.

[0021] Furthermore, the reuse information includes permission information indicating whether the second learning model generated by the first NF by the VFL is permitted to be used in the second NF. When the second learning model is permitted to be used in the second NF, the reuse information may include restriction information for limiting the second NFs permitted to use the second learning model. The restriction information may include information indicating a vendor of the second NF that can use the second learning model, or information indicating a vendor of the second NF that cannot use the second learning model. The restriction information may include information indicating a business that can use the second learning model, or information indicating a business that cannot use the second learning model. Here, the business may include a service provider or a mobile communication network operator that operates the second NF. Note that when the use of the second learning model is permitted in any second NF, it is not necessary to include restriction information in the reuse information, or restriction information indicating no usage restrictions may be included in the reuse information.

[0022] In S11, the NRF registers information indicating the first NF that sent the registration request message in association with the reuse information received in the registration request message. In S12, which is periodically or at a timing set by a mobile communication network operator, the NWDAF acquires information indicating the first NF requesting participation in VFL and the reuse information associated with the first NF from the NRF. When there are one or more first NFs requesting participation in VFL, the NWDAF selects the first NF that will perform VFL using an arbitrary method. For example, the NWDAF can select the first NF that will perform VFL using descriptive information, i.e., information indicating the analysis content, network information used as input for the first learning model, APP information used as input for the second learning model, etc.

[0023] In S13, the NWDAF notifies the first NF of the start of VFL. As a result, the NWDAF and the first NF start learning the first learning model and the second learning model. The second learning model may or may not use APP information collected by the first NF as its input. In S14, the NWDAF and the first NF end VFL. The end of VFL may be notified by the NWDAF to the first NF, or may be notified by the first NF to the NWDAF. When the first NF notifies the NWDAF of the end of VFL, the NWDAF notifies the first NF of the end of VFL as a confirmation response. Thereafter, the NWDAF sends intermediate information, which is the output of the first learning model, to the first NF, and the first NF analyzes the intermediate information from the NWDAF and the APP information collected by the first NF as input to the second learning model.

[0024] In addition, in S15, the NWDAF determines whether the second learning model generated in the first NF can be acquired based on the reuse information associated with the first NF. Specifically, if the reuse information included in the reuse information associated with the first NF indicates permission, the NWDAF determines that the second learning model can be acquired. If the reuse information indicates non-permission, the NWDAF determines that the second learning model cannot be acquired. If the second learning model cannot be acquired, the NWDAF does not perform the processes of S16 and S17 in FIG. 3. On the other hand, if the second learning model can be acquired, the NWDAF acquires the second learning model from the first NF in S16. Then, in S17, the NWDAF associates the second learning model acquired in S16 with the description information and restriction information included in the reuse information associated with the first NF acquired in S12, and stores the second learning model within the NWDAF or another NF equipped with a storage function.

[0025] Figure 4 is a sequence diagram of the process by which the second NF acquires a second learning model. It is assumed that the NWDAF stores multiple pairs of the first learning model and the second learning model generated in the sequence of Figure 3. In S20, the second NF transmits an analysis request message to the NWDAF. The analysis request message includes information indicating a request for distribution of the second learning model and analysis content, in a form in which the NWDAF uses the first learning model and the second NF performs analysis using the second learning model. In S21, the NWDAF selects a second learning model that can be distributed to the second NF from one or more second learning models that match the analysis content requested by the second NF.

[0026] The second learning model that can be distributed to the second NF is determined based on the restriction information associated with the second learning model. Therefore, although not shown in FIG. 4 , in order to determine the second learning model that can be distributed to the second NF, the NWDAF performs a process of acquiring information indicating the operator operating the second NF, information indicating the vendor of the second NF, etc., from the second NF as needed. Furthermore, the second learning model that can be distributed to the second NF must also be able to collect APP information that serves as input to the second NF. Therefore, the NWDAF performs a process of acquiring APP information that the second NF can collect from the second NF as needed. In S22, the NWDAF transmits the second learning model determined in S21 to the second NF. Thereafter, the NWDAF generates intermediate information based on the first learning model, which is a pair of the second learning model distributed to the second NF, and sends it to the second NF, and the second NF analyzes the intermediate information from the NWDAF and the APP information collected by the second NF as input to the second learning model.

[0027] As described above, the first NF can notify the NWDAF of whether or not to allow distribution of the generated second learning model to the second NF. For example, the second learning model may include APP information used for learning. Therefore, if the APP information used to learn the second learning model includes privacy information that cannot be submitted to the outside, the service provider operating the first NF can prohibit distribution of the second learning model to other NFs. On the other hand, if the APP information used to learn the second learning model does not include privacy information that cannot be submitted to the outside, or if the APP information is not used as input for the second learning model, the service provider operating the first NF can permit distribution of the second learning model to other NFs. For example, the service provider operating the first NF must pay a fee to the mobile communication network operator for using the analysis service provided by the NWDAF. Here, for example, if distribution of the second learning model is permitted, the fee can be set lower than if distribution of the second learning model is not permitted. In this case, the service provider operating the first NF can reduce the cost of using the analysis service by allowing the distribution of the second learning model that does not contain privacy information.

[0028] Furthermore, in this embodiment, even if the service provider operating the first NF permits distribution of the second learning model, the service provider operating the first NF can impose restrictions on the distribution destination. Therefore, the service provider operating the first NF can prevent the second learning model from being distributed to undesirable service providers / operators. Furthermore, the service provider operating the first NF can prevent the second learning model from being used in devices from undesirable vendors.

[0029] <Device Configuration> Fig. 5 is a diagram showing an example configuration of a network node 1 that implements NWDAF. The network node 1 has, for example, one or more processors and one or more memory devices. The one or more memory devices may include volatile memory devices and non-volatile memory devices. Each functional block shown in Fig. 5 may be realized by one or more processors executing a computer program stored in the one or more memory devices. Furthermore, the network node 1 may be realized by a single device. Alternatively, the network node 1 may be realized by multiple devices that can communicate with each other. Note that Fig. 5 shows only parts necessary for understanding the embodiment, and the network node 1 may include other functional blocks that are not shown.

[0030] The generation unit 11 performs vertical federated learning with the server 3 to generate a first learned model. If the server 3 has notified the mobile communication network that the second learned model generated by the server 3 through vertical federated learning can be transmitted to the mobile communication network, the request unit 12 requests the server 3 to transmit the second learned model after the vertical federated learning is completed. On the other hand, if the server 3 has notified the mobile communication network that the second learned model cannot be transmitted to the mobile communication network, the request unit 12 does not request the server 3 to transmit the second learned model. The notification of whether the second learned model can be transmitted to the mobile communication network is indicated in the availability information that the server 3 transmits to the mobile communication network to request participation in the vertical federated learning. Although not shown in the sequence of FIG. 3 , the server 3 (first NF) can change the content of the availability information before the vertical federated learning is completed.

[0031] When the storage unit 13 receives the second learning model from the server 3, it stores the second learning model in a storage device in association with the restriction information and description information of the reuse information notified by the server 3. The storage device may be a device inside the network node 1. Alternatively, the storage device may be a device outside the network node 1 that is accessible by the network node 1. When the distribution unit 14 receives a subscription request to a service that uses the second learning model from the server 4, it determines a second learning model that can be distributed to the server 4 based on the restriction information, etc., and distributes the determined second learning model to the server 4.

[0032] FIG. 6 is a configuration diagram of a server 3 according to some embodiments. The server 3 includes, for example, one or more processors and one or more memory devices. The one or more memory devices may include volatile memory devices and non-volatile memory devices. Each functional block shown in FIG. 6 may be realized by one or more processors executing a computer program stored in the one or more memory devices. The server 3 may also be realized by a single device. Alternatively, the server 3 may be realized by multiple devices capable of communicating with each other. Note that FIG. 6 shows only parts necessary for understanding the embodiments, and the server 3 may include other functional blocks not shown.

[0033] The transmitter 33 and the receiver 32 process the transmission and reception of messages to and from the mobile communication network. The transmitter 33 transmits a message to a network node 2 of the mobile communication network indicating participation in vertical federated learning performed with a network node 1 of the mobile communication network. The transmitter 33 includes in the message availability information indicating whether the learning model generated in the vertical federated learning can be transmitted to the mobile communication network. If the learning model can be transmitted to the mobile communication network, the transmitter 33 can include in the message limitation information for limiting the distribution destinations of the learning model.

[0034] The restriction information may be information indicating a service provider or a mobile communication network operator to which the learning model can be distributed. The restriction information may be information indicating a service provider or a mobile communication network operator to which the learning model cannot be distributed. The restriction information may be information indicating a vendor of a device (NF) that is prohibited from running the learning model. The restriction information may be information indicating a vendor of a device (NF) that is permitted to run the learning model.

[0035] The network node 1 may be a node that implements the NWDAF, and the network node 2 may be a node that implements the NRF. The generation unit 31 generates a second learning model in vertical federated learning performed with the network node 1.

[0036] The present disclosure provides a computer program, which, when executed by one or more processors of an apparatus having one or more processors, causes the apparatus to function as a network node 1 or a server 3, and a computer-readable storage medium storing the computer program. Furthermore, the present disclosure provides a method executed by the network node 1 for the processing shown in Figure 3 or Figure 4, a method executed by the server 3 for the processing shown in Figure 3, a computer program causing the network node 1 / server 3 to execute the method shown in Figure 3 or Figure 4, and a computer-readable storage medium storing the computer program.

[0037] The invention is not limited to the above-described embodiment, and various modifications and variations are possible within the scope of the gist of the invention.

[0038] This application claims priority based on Japanese Patent Application No. 2024-031585, filed March 1, 2024, the entire contents of which are incorporated herein by reference.

Claims

1. A server comprising: a transmitting means configured to transmit a first message to a second network node of a mobile communication network indicating participation in vertical federated learning with a first network node of the mobile communication network; and a generating means configured to generate a learning model in the vertical federated learning, wherein the first message includes availability information indicating whether the learning model generated in the vertical federated learning can be transmitted to the mobile communication network.

2. The server of claim 1, wherein, if the availability information indicates that the learning model can be transmitted to the mobile communication network, the first message includes limitation information for limiting the distribution destinations of the learning model.

3. The server described in claim 2, wherein the restriction information includes information indicating a business or operator to which the learning model can be distributed, information indicating a business or operator to which the learning model cannot be distributed, information indicating a vendor of a device to which the learning model is permitted to be used, or information indicating a vendor of a device to which the learning model is prohibited from being used.

4. A server according to any one of claims 1 to 3, wherein the first network node is a node implementing a Network Data Analysis Function (NWDAF).

5. A server according to any one of claims 1 to 4, wherein the second network node is a node implementing a Network Repository Function (NRF).

6. A computer-readable storage medium storing a program that, when executed by one or more processors of a device having one or more processors, causes the device to function as a server according to any one of claims 1 to 5.

7. A network node of a mobile communication network, comprising: a generation means configured to generate a first learning model through vertical federated learning performed with a server; and a request means configured to request the server to transmit the second learning model after completion of the vertical federated learning, when the server has notified the mobile communication network that it can transmit to the mobile communication network a second learning model generated by the server through the vertical federated learning.

8. A network node as described in claim 7, wherein the requesting means is configured not to request the server to transmit the second learning model if the server has notified the mobile communication network that it cannot transmit the second learning model to the mobile communication network.

9. A network node as described in claim 7 or 8, further comprising a storage means configured to store the second learning model received from the server in response to the request means requesting the server to transmit the second learning model in a storage device in association with limitation information notified by the server for limiting the distribution destinations of the second learning model.

10. A network node according to any one of claims 7 to 9, wherein the network node is a node implementing a Network Data Analysis Function (NWDAF).

11. A computer-readable storage medium storing a program that, when executed by one or more processors of a device having one or more processors, causes the device to function as a network node according to any one of claims 7 to 10.

Images