Core network for mobile communication network, network node and program

Abstract

To provide a technique which increases the number of options relating to provision of a model to three or more.SOLUTION: A core network for a mobile communication network comprises: storage means for storing a plurality of models being generated based on learning data and used for the same application but at different levels, the level of the model indicating a degree of the amount of information relating to learning data included in the model; determination means for determining, in response to receiving a message requesting a first model for a first application from a consumer node, whether or not the first model can be provided to the consumer node and, in a case where the first model can be provided to the consumer node, a level of the first model that can be provided to the consumer node; and notification means for notifying, in a case where it is determined that a first learning model at a first level can be provided to the consumer node, the consumer node of information identifying the first model of the first level.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present disclosure relates to a core network and network nodes of a mobile communication network. [Background technology]

[0002] The Third Generation Partnership Project (3GPP (registered trademark)) defines a Network Data Analysis Function (NWDAF) for analyzing the current state of a mobile communication network and inferring the future state. The NWDAF includes at least one of a Model Training Logic Function (MTLF) and an Analysis Logic Function (AnLF). The MTLF performs machine learning based on learning data collected from each Network Function (NF) and device of the mobile communication network to generate a learning model (hereinafter, simply referred to as a model). The AnLF performs inference using the model generated by the MTLF. In addition, the 3GPP (registered trademark) also defines an Analysis Data Repository Function (ADRF) for storing the analysis results or inference results of the NWDAF.

[0003] The analysis or inference results of the NWDAF may be used not only within the operator of the mobile communication network that operates the NWDAF, but also provided to NFs operated by organizations outside the operator for use in those organizations. In the following description, an NF that obtains the analysis or inference results of the NWDAF is referred to as an NF service consumer (NFc).

[0004] The analysis result or inference result of the NWDAF may include privacy information (e.g., user location information) about a user who uses a mobile communication network including the NWDAF. For this reason, 3GPP (registered trademark) is discussing changing the degree of anonymization of data showing the analysis result and inference result depending on the NFc to which the analysis result or inference result is provided.

[0005] Furthermore, a mobile communication network operator may provide the NFc with the model itself generated by the NWDAF (MTLF) operated by the operator.

[0006] Here, Non-Patent Document 1 discloses that a model generated by machine learning includes information about the training data used to train the model. Therefore, if the training data includes privacy information, the model generated by machine learning also includes privacy information. [Prior art documents] [Non-patent literature]

[0007] [Non-Patent Document 1] Nasr, Milad, Reza Shokri, and Amir Houmansadr, "Comprehensive privacy analysis of deep learning", Proceedings of IEEE Symposium on Security and Privacy, 2018. [Non-Patent Document 2] Abadi, Martin, et.al., "Deep learning with differential privacy", Proceedings of ACM SIGSAC conference on computer and communications security, 2016. Summary of the Invention [Problem to be solved by the invention]

[0008] For this reason, 3GPP (registered trademark) is currently considering providing models only to NFc selected in advance by operators, and not providing models to other NFc. Therefore, operators have only two options: to provide models to NFc or not to provide them.

[0009] The present disclosure provides techniques that provide more than two options for providing models. [Means for solving the problem]

[0010] According to one aspect of the present disclosure, a core network of a mobile communication network includes a storage means for storing multiple learning models of different levels but used for the same purpose, which are generated based on learning data, wherein the level of a learning model indicates the degree of the amount of information related to the learning data contained in the learning model; a determination means for determining, in response to receiving a message requesting a first learning model for a first purpose from a consumer node, whether the first learning model can be provided to the consumer node and, if the first learning model can be provided to the consumer node, the level of the first learning model that can be provided to the consumer node; and a notification means for notifying the consumer node of information identifying the first learning model of the first level when it is determined that the first learning model of the first level can be provided to the consumer node. Effect of the Invention

[0011] According to the present disclosure, there may be more than two options for providing the model. [Brief description of the drawings]

[0012] [Figure 1] FIG. 11 is a sequence diagram showing an example of a model generation process and a provision process according to the background art. [Diagram 2] FIG. 2 is a diagram showing determination information stored in the NRF in the sequence of FIG. 1; [Diagram 3] FIG. 1 is a diagram showing an example of the configuration of a system according to an embodiment. [Figure 4] FIG. 4 is a sequence diagram illustrating an example of a model generation process according to an embodiment. [Diagram 5] FIG. 4 is a diagram showing an example of each piece of information according to an embodiment. [Figure 6] FIG. 11 is a sequence diagram showing an example of a model providing process according to an embodiment. [Figure 7] 1 is a functional block diagram of a core network according to an embodiment. [Figure 8] 1 is a functional block diagram of a network node according to an embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, the embodiments will be described in detail with reference to the accompanying drawings. Note that the following embodiments do not limit the invention according to the claims, and not all combinations of features described in the embodiments are essential to the invention. Two or more features among the multiple features described in the embodiments may be arbitrarily combined. In addition, the same reference numbers are used for the same or similar configurations, and duplicated descriptions are omitted.

[0014] Before describing the embodiment, an example of the current model generation process and provision process will be described with reference to FIG. 1. In S1, the NWDAF generates a model based on learning data. Here, the purpose of the model generated in S1 is "A" and its identifier (ID) is "X". In S2, the NWDAF notifies the ADRF of the main body of the model generated in S1 and its identifier X, and notifies the network repository function (NRF) of the identifier X of the model generated in S1 and its purpose A. In S3, the NRF registers the relationship between the model identifier X and the purpose A of the model in the model information. In addition, the ADRF stores the model main body in association with the identifier X in S4.

[0015] In S5, NFc#1, which is one of the NFcs, transmits a request message requesting a model of use A to the NRF. The NRF determines in S6 whether the model of use A can be provided to NFc#1 based on preset determination information. FIG. 2 shows an example of the determination information. According to FIG. 2, it is shown that the model of use "A" specified by identifier X can be provided to NFc#1 and NFc#2, but cannot be provided to NFc#3. Therefore, the NRF determines in S6 that the model can be provided according to the determination information in FIG. 2, and transmits a permission message including identifier X to NFc#1 in S7. The NFc#1 transmits a request message of the model body including identifier X to the ADRF in S8. In response to the request message, the ADRF transmits the model body of identifier X to NFc#1 in S9. For example, when NFc#3 requests a model for use A, the NRF determines in S6 that the model cannot be provided based on the determination information in FIG. 2, and notifies NFc#3 that the model is not permitted in S7.

[0016] As is clear from the determination information in FIG. 2, the choice of the mobile communication network operator is to decide for each NFc whether or not to provide a model, that is, there are two options.

[0017] The present embodiment will be described below. Fig. 3 is a system configuration diagram according to the present embodiment. A core network 10 of a mobile communication network includes a management function 1, an ADRF 2, an NWDAF 3, and an NRF 4. In the following description, an operator of a mobile communication network including the core network 10 is referred to as a "interested operator".

[0018] NWDAF3 has a function of performing machine learning based on training data to generate a model. Note that NWDAF3 of this embodiment has a function of generating multiple training data of different levels for the same purpose based on training data. The level indicates the degree of the amount of information (hence, the amount of privacy information) about the training data included in the model. For example, as disclosed in Non-Patent Document 2, NWDAF3 can generate models of various levels for the same purpose based on differential privacy. According to Non-Patent Document 2, the amount of privacy information included in the model is controlled by two parameters ε and δ. Note that, in general, the less the amount of information about the training data included in the model, the more the performance of the model, for example, inference performance, deteriorates.

[0019] ADRF2 stores the model generated by NWDAF3. NRF4 functions as a window for NFc50. NFc50 is an NF operated by an external organization different from the operator of interest, and is also referred to as a consumer node. In addition, external organizations may include not only operators of other mobile communication networks different from the operator of interest, but also operators of other types of networks such as Internet service providers (ISPs), and companies and organizations that use the services of the mobile communication network operated by the operator of interest.

[0020] NFc50 has a function of acquiring a model generated by NWDAF3. Although five NFc50 are shown in Fig. 3, the number of NFc50 may be any number equal to or greater than one. In the following description, when the five NFc50 are to be distinguished, they are represented as NFc#1 to NFc#5 as shown in Fig. 3. Management function 1 controls the provision of models generated by NWDAF3 to NFc#1 to NFc#5.

[0021] FIG. 4 shows a sequence when generating a model. In S10, the management function 1 instructs the NWDAF 3 to generate a model for use A according to an input from the operator of interest. At this time, the management function 1 instructs the NWDAF 3 to generate a plurality of models with different levels according to the input from the operator of interest. In FIG. 4, the generation of three models of levels L1 to L3 is instructed. In this example, the larger the level number, the less the content of privacy information. For example, the level L1 can be a model generated by machine learning that does not consider the amount of privacy information included in the model. In this example, the management function 1 also notifies the NWDAF 3 of the identifiers of the models of each level. According to FIG. 4, the identifiers of the models of levels L1, L2, and L3 are X1, X2, and X3. In S12, the management function 1 registers the model information. Note that registering the model information means storing the model information in the management function 1, or storing the model information in another NF that the management function 1 can access. As shown in FIG. 5(A), the model information indicates the relationship between the identifiers of the three models generated for the purpose A and their levels.

[0022] In S13, the NWDAF3 generates three models for application A in response to a model generation instruction from the management function 1. In the following explanation, a model body at level Ln (n is an integer from 1 to 3) is denoted as body #Ln. In S14, the NWDAF3 transmits the generated three model bodies and information on their identifiers to the ADRF2. In S15, the ADRF2 stores the three model bodies in association with their identifiers, as shown in FIG. 5(B).

[0023] In the example of FIG. 4, models of three different levels are generated, but the number of levels can be any number equal to or greater than two. Furthermore, in the example of FIG. 4, the management function 1 notifies the NWDAF3 of the identifier of the model to be generated, but the NWDAF3 may be configured to determine the identifier of the model. In that case, in S10, the management function 1 notifies the NWDAF3 of the purpose and level of the model to be generated, and the NWDAF3 determines the identifier of the model of each level and notifies the management function 1. Then, the management function 1 registers the model information based on the identifier of the model of each level notified by the NWDAF3.

[0024] 4, the NWDAF 3 generates a model based on a generation instruction from the management function 1, but the trigger for generating a model is not limited to a generation instruction from the management function 1, such as, for example, an operator of interest directly inputting a model generation instruction to the NWDAF 3. In this case, the NWDAF 3 notifies the management function 1 of the purpose, level, and identifier of the generated model.

[0025] In addition, the management function 1 or other NFs accessible by the management function 1 stores in advance the judgment information shown in Fig. 5(C). The judgment information is information indicating whether a model can be provided for each NFc50, and if a model can be provided, the level of the model that can be provided. According to Fig. 5(C), it is shown that a model of level L1 can be provided for NFc#1, a model of level L2 can be provided for NFc#2 and NFc#5, a model of level L3 can be provided for NFc#3, and a model cannot be provided for NFc#4. Note that the judgment information may indicate only the level at which a model can be provided for NFc50s that can provide a model. In this case, NFc50s that are not included in the judgment information are treated as those for which a model cannot be provided.

[0026] The judgment information may be determined in advance, for example, based on an agreement between the noted operator and an organization that operates each NFc 50. Alternatively, the judgment information may be determined solely by the noted operator.

[0027] 6 is a sequence diagram of the model providing process. In S20, the NFc#1 transmits a request message requesting a model of use A to the NRF4. In S21, the NRF4 transmits a request message indicating that the NFc#1 is requesting a model of use A to the management function 1.

[0028] In S22, the management function 1 judges whether or not a model of application A can be provided to NFc#1, and if so, what level it can be provided at. According to the judgment information in FIG. 5(C), a model of level L1 can be provided to NFc#1. Therefore, in S23, the management function 1 notifies the NRF4 of a permission message. The permission message includes information indicating the identifier X1 of the model of level L1 of application A, and information specifying NFc#1 that is the object of permission. The management function 1 judges the identifier X1 of the model of level L1 of application A by referring to the model information.

[0029] In S24, the NRF4 sends a permission message to the NFc#1. The permission message includes information indicating the identifier X1 of the model of the level L1 of the purpose A. In S25, the NFc#1 sends a request message for the model body including information indicating the identifier X1 to the ADRF2. In response to the request message, in S26, the ADRF2 sends the body #L1 of the model with the identifier X1 to the NFc#1.

[0030] For example, when NFc#2 requests a model of use A, the management function 1 determines that a model of level L2 can be provided based on the determination information in Fig. 5(C), and transmits a permission message including information indicating identifier X2 in S23. Also, when NFc#4 requests a model of use A, the management function 1 determines that the model cannot be provided based on the determination information in Fig. 5(C), and transmits a message indicating that permission is not permitted in S23. When a model cannot be provided, the management function 1 may be configured not to transmit any message to the requesting NFc50, rather than transmitting a message indicating that permission is not permitted to the requesting NFc50.

[0031] 6, the determination result in S22 by the management function 1 is notified to the NFc 50 via the NRF 4, but the management function 1 may be configured to directly notify the NFc 50. In addition, the determination information and model information may be stored in the NRF 4, or the NRF 4 may be configured to have access to the determination information and model information stored in the management function 1 or another NF, and the NRF 4 may determine whether or not a model can be provided, and if so, the level of the model.

[0032] 5C is used in common for all uses regardless of the use of the model, but it is also possible to provide judgment information for each use of the model. In this case, for example, for a model for use A, a model of level L1 is provided to NFc#1, but for a model for use B, a model of a level different from level L1 is not provided to NFc#1.

[0033] With the above configuration, it becomes possible to select from three or more options regarding the provision of models to NFc50, rather than selecting one of two options, that is, whether or not to provide the model. In other words, it is possible to diversify the options for providing the model to three or more.

[0034] Each of the management function 1, ADRF2, NWDAF3, and NRF4 shown in Fig. 3 can be implemented as a single device or as multiple devices that can communicate with each other. Also, two or more functions of the management function 1, ADRF2, NWDAF3, and NRF4 can be implemented as a single device.

[0035] Fig. 7 is a functional block diagram of the core network 10 according to the embodiment. The generation unit 14 performs processing to generate a plurality of models used for the same purpose but at different levels based on the learning data and store them in the storage unit 11. The generation unit 14 corresponds to, for example, NWDAF3 (MTLF) in Fig. 3. The storage unit 11 corresponds to ADRF2 in Fig. 3.

[0036] In response to receiving a message requesting a first learning model for a first purpose from the NFc50, the determination unit 12 determines whether or not a first learning model can be provided to the NFc50, and if so, the level of the first learning model that can be provided. This determination can be made by referring to model information and determination information. The determination unit 12 corresponds to the management function 1 in Fig. 3. As described above, when the process of S22 in Fig. 6 is performed by the NRF4, the determination unit 12 corresponds to the NRF4.

[0037] When the determination unit 12 determines that the first learning model can be provided to the NFc50, the notification unit 13 notifies the NFc50 of information, for example, an identifier, specifying a level of the first learning model that can be provided. When the determination unit 12 determines that the first learning model cannot be provided to the NFc50, the notification unit 13 explicitly indicates that the first learning model cannot be provided to the NFc50 by sending a message, or implicitly indicates that the first learning model cannot be provided to the NFc50 by not returning a response. In the sequence of FIG. 6, the notification unit 13 corresponds to the management function 1 and the NRF4. Note that when the management function 1 directly notifies the NFc50 of the determination result, the notification unit 13 corresponds to the management function 1, and when the process of S22 in FIG. 6 is performed by the NRF4, the notification unit 13 corresponds to the NRF4.

[0038] 8 is a configuration diagram of a network node 20 of a mobile communication network according to the present embodiment. The network node 20 may be, for example, a device of the core network 10. The network node 20 may be, for example, a device that implements the management function 1 of FIG. 3, a device that implements the NRF4, or a device that implements both the management function 1 and the NRF4.

[0039] In response to receiving a message requesting a first learning model for a first purpose from the NFc50, the determination unit 21 determines whether or not a first learning model can be provided to the NFc50, and if so, the level of the first learning model that can be provided. This determination can be made by referring to model information and determination information. Note that the model information and determination information can be stored in the determination unit 12. Alternatively, one or both of the model information and the determination information can be stored in another device different from the network node 20. In this case, the determination unit 12 makes the determination by referring to the information stored in the other device.

[0040] The notification unit 22 performs processing for notifying the NFc50 of the determination result by the determination unit 21. When the determination unit 21 determines that the first learning model can be provided to the NFc50, the determination result includes information, for example, an identifier, that identifies the first learning model at a level that can be provided. The processing for notifying the NFc50 of the determination result may be, for example, processing for directly transmitting a message indicating the determination result to the NFc50. Alternatively, the processing for notifying the NFc50 of the determination result may be processing for notifying another device of the determination result and notifying the NFc50 of the determination result via the other device.

[0041] The network node 20 according to the present disclosure can be realized by a computer program that, when executed by one or more processors of a device having one or more processors, causes the device to operate as the network node 20. These computer programs can be stored in a computer-readable storage medium or distributed via a network.

[0042] According to this embodiment, the method of providing models can be diversified. Therefore, it is possible to contribute to Goal 9 of the Sustainable Development Goals (SDGs) led by the United Nations, "Build resilient infrastructure, promote sustainable industrialization and foster innovation."

[0043] 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. [Explanation of symbols]

[0044] 11: storage unit, 12: judgment unit, 13: notification unit

Claims

1. A core network of a mobile communication network, A storage means for storing a plurality of learning models, which are generated based on learning data and are used for the same purpose but have different levels, the level of the learning model indicating the degree of the amount of information related to the learning data included in the learning model; a determination means for determining, in response to receiving a message from a consumer node requesting a first learning model for a first use, whether the first learning model can be provided to the consumer node, and, if the first learning model can be provided to the consumer node, the level of the first learning model that can be provided to the consumer node; a notification means for notifying the consumer node of information identifying the first learning model of the first level when it is determined that the first learning model of the first level can be provided to the consumer node; A core network that has

2. The core network of claim 1, wherein when the determination means determines that the first learning model cannot be provided to the consumer node, the notification means notifies the consumer node that the first learning model cannot be provided, or does not return a response to the message to the consumer node.

3. The core network described in claim 1, wherein the determination means determines whether the first learning model can be provided to the consumer node based on determination information indicating the level of the learning model that can be provided to the consumer node, and if the first learning model can be provided to the consumer node, the level of the first learning model that can be provided to the consumer node.

4. The core network according to claim 3 , wherein the determination information is information that is commonly used regardless of the application of the learning model.

5. The determination information is provided for each of the applications of the learning model, The core network described in claim 3, wherein the determination means determines whether the first learning model can be provided to the consumer node based on the determination information of the first use, and if the first learning model can be provided to the consumer node, the level of the first learning model that can be provided to the consumer node.

6. The core network according to claim 1 , further comprising a generating means for generating the plurality of learning models used for the same purpose but having different levels based on the learning data and storing the models in the storage means.

7. The core network of claim 1 , wherein the consumer nodes are network nodes operated by an organization different from an operator of the core network.

8. A network node of a mobile communication network, comprising: The mobile communication network stores a plurality of learning models, which are generated based on learning data and are used for the same purpose but have different levels, the levels of the learning models indicating a degree of an amount of information related to the learning data included in the learning models; The network node: a determination means for determining, in response to receiving a message from a consumer node requesting a first learning model for a first use, whether the first learning model can be provided to the consumer node, and, if the first learning model can be provided to the consumer node, the level of the first learning model that can be provided to the consumer node; a notification means for performing a process of notifying the consumer node of information identifying the first learning model of the first level when it is determined that the first learning model of the first level can be provided to the consumer node; A network node comprising:

9. The network node described in claim 8, wherein the determination means determines whether the first learning model can be provided to the consumer node based on determination information indicating the level of the learning model that can be provided to the consumer node, and if the first learning model can be provided to the consumer node, the level of the first learning model that can be provided to the consumer node.

10. A program that, when executed by one or more processors of an apparatus having one or more processors, causes the apparatus to function as a network node according to claim 8 or 9.

Images