Support of sample id intersection for vfl

Abstract

The disclosure relates to a 5G or 6G communication system for supporting a higher data transmission rate. There is provided a method performed by a network data analytics function (NWDAF) performing vertical federated learning (VFL) in a wireless communication system, the method comprising: transmitting, to a network repository function (NRF), a registration message including a network function (NF) profile of the NWDAF, wherein a capability of aggregating at least one intermediate result for at least one VFL of other NWDAF is included in the NF profile; receiving, from the NRF, a registration response corresponding to the registration message; receiving, from an application function (AF) via a network exposure function (NEF), a VFL preparation request message; transmitting, to at least one other NWDAF, at least one VFL preparation request message; receiving, from the at least one other NWDAF, at least one response from at least one other NWDAF, based on the at least one VFL preparation request message; aggregating the at least one received response; and transmitting, to the AF via the NEF, the aggregated response.

Description

SUPPORT OF SAMPLE ID INTERSECTION FOR VFL

[0001] Certain examples of the present disclosure provide various techniques relating to Virtual Federated Learning (VFL) for example within 3rdGeneration Partnership Project (3GPP) 5th Generation (5G) New Radio (NR) networks.

[0002]

[0003] 5G mobile communication technologies define broad frequency bands such that high transmission rates and new services are possible, and can be implemented not only in "Sub 6GHz" bands such as 3.5GHz, but also in "Above 6GHz" bands referred to as mmWave including 28GHz and 39GHz. In addition, it has been considered to implement 6G mobile communication technologies (referred to as Beyond 5G systems) in terahertz bands (for example, 95GHz to 3THz bands) in order to accomplish transmission rates fifty times faster than 5G mobile communication technologies and ultra-low latencies one-tenth of 5G mobile communication technologies.

[0004] At the beginning of the development of 5G mobile communication technologies, in order to support services and to satisfy performance requirements in connection with enhanced Mobile BroadBand (eMBB), Ultra Reliable Low Latency Communications (URLLC), and massive Machine-Type Communications (mMTC), there has been ongoing standardization regarding beamforming and massive MIMO for mitigating radio-wave path loss and increasing radio-wave transmission distances in mmWave, supporting numerologies (for example, operating multiple subcarrier spacings) for efficiently utilizing mmWave resources and dynamic operation of slot formats, initial access technologies for supporting multi-beam transmission and broadbands, definition and operation of BWP (BandWidth Part), new channel coding methods such as a LDPC (Low Density Parity Check) code for large amount of data transmission and a polar code for highly reliable transmission of control information, L2 pre-processing, and network slicing for providing a dedicated network specialized to a specific service.

[0005] Currently, there are ongoing discussions regarding improvement and performance enhancement of initial 5G mobile communication technologies in view of services to be supported by 5G mobile communication technologies, and there has been physical layer standardization regarding technologies such as V2X (Vehicle-to-everything) for aiding driving determination by autonomous vehicles based on information regarding positions and states of vehicles transmitted by the vehicles and for enhancing user convenience, NR-U (New Radio Unlicensed) aimed at system operations conforming to various regulation-related requirements in unlicensed bands, NR UE Power Saving, Non-Terrestrial Network (NTN) which is UE-satellite direct communication for providing coverage in an area in which communication with terrestrial networks is unavailable, and positioning.

[0006] Moreover, there has been ongoing standardization in air interface architecture / protocol regarding technologies such as Industrial Internet of Things (IIoT) for supporting new services through interworking and convergence with other industries, IAB (Integrated Access and Backhaul) for providing a node for network service area expansion by supporting a wireless backhaul link and an access link in an integrated manner, mobility enhancement including conditional handover and DAPS (Dual Active Protocol Stack) handover, and two-step random access for simplifying random access procedures (2-step RACH for NR). There also has been ongoing standardization in system architecture / service regarding a 5G baseline architecture (for example, service based architecture or service based interface) for combining Network Functions Virtualization (NFV) and Software-Defined Networking (SDN) technologies, and Mobile Edge Computing (MEC) for receiving services based on UE positions.

[0007] As 5G mobile communication systems are commercialized, connected devices that have been exponentially increasing will be connected to communication networks, and it is accordingly expected that enhanced functions and performances of 5G mobile communication systems and integrated operations of connected devices will be necessary. To this end, new research is scheduled in connection with eXtended Reality (XR) for efficiently supporting AR (Augmented Reality), VR (Virtual Reality), MR (Mixed Reality) and the like, 5G performance improvement and complexity reduction by utilizing Artificial Intelligence (AI) and Machine Learning (ML), AI service support, metaverse service support, and drone communication.

[0008] Furthermore, such development of 5G mobile communication systems will serve as a basis for developing not only new waveforms for providing coverage in terahertz bands of 6G mobile communication technologies, multi-antenna transmission technologies such as Full Dimensional MIMO (FD-MIMO), array antennas and large-scale antennas, metamaterial-based lenses and antennas for improving coverage of terahertz band signals, high-dimensional space multiplexing technology using OAM (Orbital Angular Momentum), and RIS (Reconfigurable Intelligent Surface), but also full-duplex technology for increasing frequency efficiency of 6G mobile communication technologies and improving system networks, AI-based communication technology for implementing system optimization by utilizing satellites and AI (Artificial Intelligence) from the design stage and internalizing end-to-end AI support functions, and next-generation distributed computing technology for implementing services at levels of complexity exceeding the limit of UE operation capability by utilizing ultra-high-performance communication and computing resources.

[0009] In R19 SA2 AIML_CN topic, VFL (vertical federated learning) was specified by introducing main procedures for VFL process, including VFL preparation, VFL model training, VFL inference, VFL performance / accuracy monitoring etc. However, due to limited time in R19, there are many valid scenarios were dropped. Based on the current discussion on potential Rel-20 5GA topics, the proposals on supporting VFL enhancements including dynamic negotiation of Features (depending on the exception sheet progress), and change of VFL clients during training are on the table for further determination.

[0010] In the following clauses, we will review of R19 AIML_CN normative work on VFL

[0011] General description of VFL in 3GPP specifications

[0012] Vertical Federated learning is a machine learning technique working without exchanging / sharing of local data set, while maintaining some level of coordination amongst VFL participants, when training and inference are performed on local ML Models, wherein the local data set in different VFL Participant for local model training have different feature spaces for the same samples (e.g. UE IDs). Vertical Federated Learning may involve multiple NWDAFs and AFs.

[0013] For Vertical Federated Learning, there may be one NWDAF or one AF acting as a VFL server and one or multiple NWDAF(s) and / or one or multiple AF(s) acting as VFL Client(s). Vertical Federated Learning is available among NWDAFs or between NWDAF(s) and AF(s) within a single PLMN or between an AF and NWDAF(s) in a single PLMN. When AF is acting as VFL Server, NWDAF(s) is VFL Client(s).

[0014] VFL server:

[0015] -　　　 An NWDAF or trusted AF acting as VFL server discovers and selects VFL client(s) (NWDAF(s) and / or AF(s)) to participate in a VFL procedure.

[0016] NOTE:　　　 When an untrusted AF is acting as VFL server, NEF discovers and selects candidate VFL client NWDAFs, then the AF determines final set of VFL clients.

[0017] -　　　 It requests VFL clients to do local ML model training for an Analytic ID, it assigns VFL correlation ID, and it requests to report intermediate results.

[0018] -　　　 It optionally locally trains ML Model with the available local data set.

[0019] -　　　 It combines intermediate results from VFL client(s) and VFL server and computes intermediate training results (e.g. gradient information, loss information) for updating its own local ML Model and the ML Models of VFL clients during the VFL training process and sends the intermediate training results towards VFL clients involved in the joint VFL training process. VFL server may send and receive separate message for each client.

[0020] -　　　 It determines to terminate the VFL training process.

[0021] -　　　 It stores VFL correlation ID and locally trained ML Model after VFL training process.

[0022] -　　　 It initiates the VFL inference process using VFL correlation ID.

[0023] -　　　 It combines local inference result from VFL clients and generates the final VFL inference result.

[0024] -　　　 It may send the final VFL inference result to the consumer.

[0025] -　　　 It supports to monitor the accuracy of the VFL model.

[0026] VFL client:

[0027] -　　　 It locally trains ML Model with the available local data set, which includes the data that may not be allowed to be shared with other VFL clients or VFL server due to e.g. data privacy, data security, data access rights.

[0028] -　　　 It computes the intermediate results for their local ML Models involved in the VFL training and provide reports with the intermediate results to the AF or NWDAF acting as VFL server.

[0029] -　　　 It stores VFL correlation ID and locally trained ML model after VFL training process.

[0030] -　　　 It performs inference based on the local model and local data and provides inference results to VFL server.

[0031] Vertical Federated Learning includes the following procedures:

[0032] -　　　 Registration of the NF profile including a list of VFL related information to NRF. Registration of the NWDAF profile to NRF is described in clause　5.2. Registration of the AF profile to NRF is described in clause　5.5. The procedure for registration and discovery of VFL server and VFL client is described in clause　6.2H.2.1 of TS　23.288.

[0033] -　　　 Preparation for VFL including sample alignment to ensure that all the VFL participants have common samples when training ML models as described in clause　6.2H.2.2 of TS　23.288.

[0034] -　　　 Training for VFL as described in clause　6.2H.2.3 of TS　23.288.

[0035] -　　　 Inference for VFL as described in clause　6.2H.2.4 of TS　23.288.

[0036] The following new terms have been introduced to TS　23.288 to support VFL:

[0037] ●Vertical Federated Learning (VFL):a federated learning technique without exchanging / sharing local data set, wherein the local data set in different VFL Participant for local model training have different feature spaces for the same samples (e.g. UE IDs).

[0038] ●Label: A label is the training objective in supervised machine learning.

[0039] VFL server and clients discovery and selection

[0040] As documented in TS　23.288: When selecting an NWDAF that supports Vertical Federated Learning (VFL), the following additional factors may be considered by the NWDAF:

[0041] -　　　 Time Period of Interest: time interval [start...end], during which the Vertical Federated Learning will be performed.

[0042] -　　　 when selecting VFL client NWDAF:

[0043] VFL capability type as VFL client NWDAF per Analytics ID.

[0044] VFL Interoperability Indicator per Analytics ID.

[0045] optionally, supported Feature IDs per Analytics ID.

[0046] optionally NF set ID(s) of the data source(s).

[0047] optionally the Serving Area information.

[0048] -　　　 when selecting VFL server NWDAF:

[0049] VFL capability type as VFL server NWDAF per Analytics ID.

[0050] For (untrusted and / or trusted) AF and the NEF that serves the untrusted AF, other than the existing information specified in the NF profile, they may register to NRF by including the following information in the NF profile to support VFL:

[0051] -　　　 For trusted AF as VFL client, the following information is supported:

[0052] Analytics ID(s).

[0053] VFL capability information per analytics ID including VFL capability type (i.e. VFL client),.

[0054] VFL interoperability indicator per Analytics ID

[0055] Optionally, the time interval supporting VFL.

[0056] optionally supported feature ID(s) per Analytics ID.

[0057] -　　　 For NEFs serving untrusted AFs:

[0058] Event ID(s) supported by AFs.

[0059] For VFL Analytics ID(s) supported by AFs.

[0060] VFL capability information per analytics ID including VFL capability type (i.e. VFL client), supported by AFs.

[0061] VFL interoperability indicator per Analytics ID

[0062] For VFL optionally, the time interval supporting VFL.

[0063] For VFL, optionally feature ID(s) per Analytics ID supported by AFs.

[0064] Registration and Discovery procedure for Vertical Federated Learning

[0065] FIG. 1 illustrates a registration and discovery procedure for VFL when NWDAF or trusted AF is acting as the VFL server.

[0066] Steps 1 to 3 are the NWDAF and AF Registration procedures when the VFL server is NWDAF or a trusted AF.

[0067] 1a.　VFL Server NWDAF / trusted AF registers to NRF with its NF profile, which includes NF Type (i.e. NWDAF type or AF type), Analytics ID(s), service area if available, VFL capability information per analytics ID and VFL interoperability indicator(s) and optional supported feature ID(s), and other parameters as described in clause　5.2 of TS　23.288.

[0068] 1b.　NWDAF as VFL client registers to NRF with its NF profile, which includes NF Type (i.e. NWDAF type), Analytics ID(s), service area if available, VFL capability information per analytics ID and VFL interoperability indicator(s) and optional supported feature ID(s), as defined in clause　5.2.

[0069] 1c.　When untrusted AF as VFL client, it shall register to the NEF via OAM configuration: Analytics ID(s) and its VFL capability information per supported analytics ID and VFL interoperability indicator(s) and optional supported feature ID(s), and other parameters, as defined in clause　5.5 of TS　23.288. Then NEF updates NEF profile to NRF including associated AF ID and AF's parameters described in clause 5.5 of TS　23.288.

[0070] 1d.　When trusted AF as VFL client, it registers to NRF with its NF profile, which includes NF Type (i.e. AF type), analytics ID(s), service area if available, VFL capability information per analytics ID and VFL interoperability indicator and optional supported feature IDs, and other parameters, as defined in clause　5.5 of TS　23.288.

[0071] 2.　　 The NRF receives the registrations from VFL server and VFL client(s), and stores their NF profile.

[0072] 3.　　 The NRF sends registration response to VFL server and VFL client(s).

[0073] Steps 4 to 6 are the NWDAF, AF and NEF Discovery procedures when the VFL server is NWDAF or trusted AF.

[0074] 4-6.　　 The VFL server determines that the ML Model requires VFL based on e.g. operator policy, Analytics ID, VFL interoperability　indicator(s)　and Service Area.

[0075] NOTE:　　　 Step　4 in FIG. 1 may be triggered at the VFL Server by itself or a request from a consumer.

[0076] If an NWDAF can not perform as VFL Server, it first discovers and selects another VFL Server from NRF by invoking the Nnrf_NFDiscovery_Request service operation. The following criteria might be used: Analytics ID, VFL capability type as VFL server, Time Period of Interest and optional Service Area.

[0077] Once the VFL Server is determined, the VFL Server discovers other NWDAF(s) and / or AF(s) as VFL Client from NRF by invoking the Nnrf_NFDiscovery_Request service operation. The following criteria might be used: NF type(s) (i.e. NWDAF type, AF type, or NEF type), Analytics ID, VFL capability type (i.e.VFL client), VFL interoperability indicator(s), Time Period of Interest, optional feature ID(s) and optional Service Area. The AF(s) discovered by the VFL Server may be trusted and / or untrusted, the NF type may be AF type when the AF(s) as VFL Client are trusted, and the NF type may be NEF type when the AF(s) as VFL Client are untrusted, and the NF type may contain both AF type and NEF type if both trusted AF and untrusted AF are involved as VFL clients.

[0078] When an AF is acting as a VFL server, only NWDAF(s) are discovered as VFL clients.

[0079] Registration and Discovery procedure for Vertical Federated Learning when untrusted AF is acting as the VFL server

[0080] FIG. 2 illustrates a registration and discovery procedure for VFL when an untrusted AF is acting as VFL server and NWDAF(s) are the VFL clients.

[0081] The procedure illustrated in FIG. 2 shows registration and discovery for VFL training and inference when the untrusted AF is the VFL server. There can be multiple NWDAFs as VFL clients.

[0082] NOTE　1:　 For a deployment scenario that untrusted AF is the VFL server and only one NWDAF will be an VFL client, it is assumed that VFL client information for an Analytic ID is configured in NEF and step 6 and 7 may be skipped in FIG. 2.

[0083] Steps 1 to 3 are the NWDAF and AF Registration procedures when the VFL server is untrusted AF.

[0084] 1.　　 Same as the step　1b, in FIG. 1, NWDAF as VFL client registers to NRF with its NF profile, it may include those analytics IDs on which it supports to do VFL with AF as VFL server.

[0085] NOTE　1:　 The AF can use non-standardized values of the Analytics ID. The non-standardized values can be used by the AF as the VFL server to initiate the VFL training or VFL inference and these values need to be supported by NWDAFs acting as VFL clients. It is the operator´s responsibility to guarantee that non-standardized Analytics ID values within a PLMN are unique.

[0086] 2-3.　　 Same as the steps 2-3, in FIG. 1. Based on operator local policies, the AF may register its capability as VFL server for an Analytics ID to NEF via OAM configuration, this is used to make it possible for NWDAF to trigger the AF to train a model. The AF may also register an AnalyticsID in NRF e.g. when a trained model is available.

[0087] Steps 4-10 are the NWDAF Discovery procedures when the VFL server is an untrusted AF.

[0088] 4.　　 Untrusted AF acting as the VFL server determines that VFL operations are required and the NWDAF(s) as VFL client(s) are required. The AF sends a Nnef_NFDiscovery_Request for the VFL client(s) to the NEF and for each client provides selection criteria: Analytics ID, required NF type (i.e. NWDAF type), VFL capability type (i.e. VFL client), VFL interoperability indicator(s), Time Period of Interest, optional required feature ID(s), and optional Service Area.

[0089] NOTE:　　　 Step　4 in Figure FIG. 2 may be triggered at the VFL Server by itself or when requested from a consumer.

[0090] 5.　　 The NEF checks based on configured policies whether the AF is entitled to request single or multiple VFL client(s) for the Analytics ID.

[0091] 6-7.　　 The NEF discovers VFL client(s) (i.e. NWDAF(s)) on behalf of the AF from the NRF by invoking the Nnrf_NFDiscovery_Request using the selection criteria provided by the AF as defined in step　4.

[0092] 8.　　 The NEF selects NWDAF(s), e.g., by using the NWDAF(s) NF profile and parameters including load / priority / capacity that are capable of acting as VFL client(s) and matching the received selection criteria such as location information. The NEF anonymizes NWDAF instances ID(s) and assigns external NWDAF ID(s) for each selected NWDAF instance as VFL client. The NEF stores the external NWDAF ID(s) together with information how to reach the NWDAFs.

[0093] NOTE　2:　 The external NWDAF ID assigned by NEF is temporary and can be released if needed.

[0094] 9.　　 The NEF sends Nnef_NFDiscovery_Request response only including the external NWDAF ID(s) as selected in step 8 and information how they relate to corresponding selection criteria provided by the AF, VFL interoperability indicator(s),　Time Period of Interest, optional feature ID(s) and optional Service Area to the untrusted AF.

[0095] 10.　The AF stores the received external NWDAF ID(s) and uses it subsequent interactions with the NEF to indicate the target VFL client.

[0096] NOTE　3:　 If the VFL client NWDAFs are statically preconfigured to the NEF and untrusted AF and not released, the discovery procedure can be skipped.

[0097] The AF may indicate to the NEF that the AF will no longer use the VFL process identified by VFL Correlation ID (external NWDAF IDs are no longer required). If the NEF receives this indication, it shall remove the stored external NWDAF ID(s) associated with the VFL correlation ID allocated to the AF. Then, the NEF responds to the AF.

[0098] VFL preparation procedure

[0099] The VFL preparation procedure is mainly documented in TS　23.288.

[0100] Preparation procedure for Vertical Federated Learning

[0101] General

[0102] The preparation procedure is used to check if the VFL Client(s) can meet the ML Model training requirement. The procedure includes the negotiation, between server and client(s) to enable interoperability, sample alignment and may include feature negotiation if the VFL Server did not learn the supported FeatureIDs from each VFL Client using the discovery phase, alternatively the VFL Server may know the supported FeatureIDs by a VFL Client based on configuration. The Vertical Federated Learning preparation procedure can be skipped if the VFL Server can decide which VFL Client(s) support the VFL procedure to be performed, e.g. based on local configuration or offline procedures.

[0103] FIG. 3 illustrates preparation procedure for VFL when NWDAF / Trusted AF is the VFL Server.

[0104] 1.　　 An NWDAF as VFL Server may send a Vertical Federated Learning preparation request including the Analytics ID to each of the NWDAF VFL Client(s), using Nnwdaf_VFLTraining_Request and to each of the AF VFL Clients(s), using Naf_VFLTraining_Request possibly via NEF when the VFL Client is an untrusted AF. An AF as VFL Server may send a Vertical Federated Learning preparation request including the Analytics ID to each of the NWDAF VFL Client(s), using Nnwdaf_VFLTraining_Request. The NWDAF or trusted AF as a VFL Server also provides, the suggested VFL Interoperability Information to negotiate the intermediate results that will be used in training, the suggested list of sample IDs that will be used in training, and optionally, as additional criteria for sample alignment, time window of the data samples, and required minimum sample size. When a Trusted AF is acting as a VFL Server, the VFL Client can only be an NWDAF.

[0105] 2.　　 Each VFL Client checks if it can meet the ML Model training requirement. Each VFL Client checks the list of sample IDs and required criteria for sample alignment suggested by the VFL Server, and then provides to the VFL Server the list of sample IDs that it can accept out of the sample IDs suggested by the VFL Server and satisfying the required criteria for sample alignment. Each VFL Client checks the VFL Interoperability Information and determines which VFL Interoperability information that the VFL Client supports. The VFL Clients provides the list of supported Feature IDs, which is associated to the VFL Interoperability information, to the VFL Server, if available, or VFL Server may know the supported FeatureIDs for a VFL Client based on configuration.

[0106] 3.　　 Each NWDAF VFL Client invokes Nnwdaf_VFLTraining_Response or each AF VFL Client invokes Naf_VFLTrainingRequest_Response, possibly via NEF when the AF is untrusted, to indicate to the VFL Server whether it accepts the ML Model training requirements, the VFL Client can also indicate that it cannot join the FL process.

[0107] 4.　　 The VFL server determines the final list of samples considering the samples that all selected VFL clients support, if used the Feature ID per VFL client and VFL Interoperability Information to be used for training and provide them to the selected VFL Clients at the start of the training phase, as described in clause　6.2H.2.3.1 of TS　23.288.

[0108] Preparation procedure for Vertical Federated Learning when untrusted AF is the VFL server

[0109] This section specifies the preparation (including sample alignment) procedure for untrusted AF-initiated VFL scenarios between AF and NWDAF(s) within a single PLMN.

[0110] FIG. 4 illustrates preparation procedure for Vertical Federated Learning when untrusted AF is the VFL Server.

[0111] Editor´s note:　　　 　　 For the VFL preparation procedure, whether and how NEF does pre-work of sample IDs intersection before the VFL server determines the final sample IDs is FFS.

[0112] 1-　　 The untrusted AF as VFL server sends VFL preparation requests to each of the candidate VFL client(s), using Nnef_VFLTrainingRequest_Request to check if the VFL client(s) can meet the ML Model training requirement (which includes the Analytics ID, list of sample IDs, and optionally Feature ID, etc.). The external NWDAF IDs obtained in the discovery procedure (see FIG. 1) is included to indicate the target NWDAFs. The suggested VFL Interoperability Information to negotiate the intermediate results that will be used in training.

[0113] 2-　　 The NEF maps the external NWDAF and GPSI(s) to the internal NWDAF and SUPI(s). The NEF send VFL preparation request to the corresponding candidate internal NWDAF ID using Nnwdaf_VFLTraining_Request service with the same information as provided in step　1 in FIG. 1.

[0114] 3-　　 Same as step　2 in FIG. 1.

[0115] 4-　　 Same as step　3 in clause　FIG. 1.

[0116] 5-　　 The NEF maps the internal NWDAF and SUPI(s) to the external NWDAF and GPSI(s). The NEF sends the Nnef_VFLTraining_Request Response to the VFL Server with the same information as provided in step　4.

[0117] 6-　　 Same as step　4 in FIG. 1.

[0118] The above information is presented as background information only to assist with an understanding of the present disclosure. No determination has been made, and no assertion is made, as to whether any of the above might be applicable as prior art with regard to the present invention.

[0119]

[0120] When untrusted AF is acting as the VFL server, as it has been documented in clause 6.2H.2.2.2 of TS 23.288 (see FIG. 4), the NEF will forward the VFL preparation response from every VFL client individually to the server by mapping a list of internal sample IDs to external IDs. Considering the potential high number of VFL clients and the sample IDs in each VFL preparation response, NEF will have very high workload on sample ID mapping and VFL preparation response exposure.

[0121]

[0122] It is an aim of certain examples of the present disclosure to address, solve and / or mitigate, at least partly, at least one of the problems and / or disadvantages associated with the related art, for example at least one of the problems and / or disadvantages described herein. It is an aim of certain examples of the present disclosure to provide at least one advantage over the related art, for example at least one of the advantages described herein.

[0123] Embodiments or examples disclosed in the description and / or figures falling outside the scope of the claims are to be understood as examples useful for understanding the present invention.

[0124] Other aspects, advantages, and salient features of the invention will become apparent to those skilled in the art from the following detailed description taken in conjunction with the accompanying drawings.

[0125] Embodiments of the present disclosure provide a method performed by a network data analytics function (NWDAF) performing vertical federated learning (VFL) in a wireless communication system, comprising: transmitting, to a network repository function (NRF), a registration message including a network function (NF) profile of the NWDAF, wherein a capability of aggregating at least one intermediate result for at least one VFL of other NWDAF is included in the NF profile; receiving, from the NRF, a registration response corresponding to the registration message; receiving, from an application function (AF) via a network exposure function (NEF), a VFL preparation request message; transmitting, to at least one other NWDAF, at least one VFL preparation request message; receiving, from the at least one other NWDAF, at least one response from at least one other NWDAF, based on the at least one VFL preparation request message; aggregating the at least one received response; and transmitting, to the AF via the NEF, the aggregated response.

[0126] Embodiments of the present disclosure provide a method, wherein an identity (ID) of the NWDAF selected as a VFL client is anonymized by the NEF and an external NWDAF ID of the NWDAF selected as a VFL client is assigned by the NEF; and wherein the external NWDAF ID of the NWDAF is stored, by the NEF, together with information associated with reaching the NWDAF.

[0127] Embodiments of the present disclosure provide a method, further comprising; receiving the VFL preparation request message, wherein the external NWDAF ID is mapped to an internal NWDAF ID of the NWDAF by the NEF and a generic public subscription identifier (GPSI) of a terminal is mapped to a subscription permanent identifier (SUPI) of the terminal by the NEF, based on a reception of the VFL preparation request message; and transmitting the aggregated response, wherein the internal NWDAF ID of the NWDAF is mapped to the external NWDAF ID of the NWDAF, by the NEF, based on a transmission of the aggregated response.

[0128] Embodiments of the present disclosure provide a method, wherein a request for discovering at least one NWDAF is transmitted, from the AF to the NEF; wherein the at least one NWDAF is selected, by the NEF, according to at least one criterion for selecting at least one NWDAF; and wherein the at least one criterion is provided, to the NEF by the AF and is associated with the capability of aggregating.

[0129] Embodiments of the present disclosure provide a network data analytics function (NWDAF) entity in a wireless communication system, the NWDAF entity comprising: at least one transceiver; at least one processor communicatively coupled to the at least one transceiver; and at least one memory, communicatively coupled to the at least one processor, storing instructions executable by the at least one processor individually or in any combination to cause the NWDAF entity to: transmit, to a network repository function (NRF), a registration message including a network function (NF) profile of the NWDAF, wherein a capability of aggregating at least one intermediate result for at least one VFL of other NWDAF is included in the NF profile; receive, from the NRF, a registration response corresponding to the registration message; receive, from an application function (AF) via a network exposure function (NEF), a VFL preparation request message; transmit, to at least one other NWDAF, at least one VFL preparation request message; receive, from the at least one other NWDAF, at least one response, based on the at least one VFL preparation request message; aggregate the at least one received response; and transmit, to the AF via the NEF, the aggregated response.

[0130] In examples of the present disclosure, an identity (ID) of the NWDAF selected as a VFL client may be anonymized by the NEF and an external NWDAF ID of the NWDAF selected as a VFL client may be assigned by the NEF; and the external NWDAF ID of the NWDAF may be stored, by the NEF, together with information associated with reaching the NWDAF.

[0131] In examples of the present disclosure, the NWDAF entity may be caused to; receive the VFL preparation request message, wherein the external NWDAF ID is mapped to an internal NWDAF ID of the NWDAF by the NEF and a generic public subscription identifier (GPSI) of a terminal is mapped to a subscription permanent identifier (SUPI) of the terminal by the NEF, based on a reception of the VFL preparation request message; and transmit the aggregated response, wherein the internal NWDAF ID of the NWDAF is mapped to the external NWDAF ID of the NWDAF, by the NEF, based on a transmission of the aggregated response.

[0132] In examples of the present disclosure, a request for discovering at least one NWDAF may be transmitted, from the AF to the NEF; wherein the at least one NWDAF may be selected, by the NEF, according to at least one criterion for selecting at least one NWDAF; and wherein the at least one criterion may be provided, to the NEF by the AF and may be associated with the capability of aggregating.

[0133] Embodiments of the present disclosure provide a method performed by an application function (AF) performing vertical federated learning (VFL) in a wireless communication system, the method comprising: transmitting, to a network repository function (NRF) via a network exposure function (NEF), a request message for discovering at least one network data analytics function (NWDAF), wherein at least one network function (NF) profile of at least one NWDAF is registered to the NRF; receiving, from the NRF via the NEF, a response message corresponding to the request message including an external NWDAF ID of the discovered NWDAF, based on the request message; storing the external NWDAF ID included in the response message, wherein the discovered NWDAF is selected by the NEF for the VFL; transmitting, to the discovered NWDAF via the NEF, a VFL preparation request message; and receiving, from the NWDAF via the NEF, a response message corresponding to the VFL preparation request message, wherein the response message is an aggregate of at least one response message for preparing the VFL transmitted from at least one other NWDAF to the discovered NWDAF.

[0134] In examples of the present disclosure, transmitting a request message for discovering at least one NWDAF may further comprise; transmitting, to the NEF, a message for registering capability as a VFL server for an Analytics ID; and transmitting the request message for discovering at least one NWDAF.

[0135] In examples of the present disclosure, the method may further comprise receiving, from the NWDAF via the NEF, a response message corresponding to the VFL preparation request message; wherein a VFL preparation request message may be transmitted, from the discovered NWDAF to the at least one other NWDAF and a response corresponding to the request may be transmitted, from the at least one other NWDAF to the discovered NWDAF and may be aggregated by the discovered NWDAF.

[0136] In examples of the present disclosure, the method may further comprise transmitting, to the discovered NWDAF via the NEF, a VFL preparation request message, wherein the external NWDAF ID of the discovered NWDAF may be mapped to an internal NWDAF ID of the discovered NWDAF by the NEF and a generic public subscription identifier (GPSI) of a terminal may be mapped to a subscription permanent identifier (SUPI) of the terminal by the NEF; and receiving, from the discovered NWDAF via the NEF, a response message corresponding to the VFL preparation request message, wherein the internal NWDAF ID of the discovered NWDAF may be mapped to the external NWDAF ID of the discovered NWDAF, by the NEF.

[0137] Embodiments of the present disclosure provide an application function (AF) entity in a wireless communication system, the AF entity comprising: at least one transceiver; at least one processor communicatively coupled to the at least one transceiver; and at least one memory, communicatively coupled to the at least one processor, storing instructions executable by the at least one processor individually or in any combination to cause the AF entity to: transmit, to a network repository function (NRF) via a network exposure function (NEF), a request message for discovering at least one network data analytics function (NWDAF), wherein the at least one network function (NF) profile of at least one NWDAF is registered to the NRF; receive, from the NRF via the NEF, a response message corresponding to the request message including an external NWDAF ID of the discovered NWDAF, based on the request message; store the external NWDAF ID included in the response message, wherein the discovered NWDAF is selected by the NEF for the VFL; transmit, to the discovered NWDAF via the NEF, a VFL preparation request message; and receive, from the NWDAF via the NEF, a response message corresponding to the VFL preparation request message, the response message is an aggregate of at least one response message for preparing the VFL transmitted from at least one other NWDAF to the discovered NWDAF.

[0138] In examples of the present disclosure, the AF entity may be further caused to; transmit, to the NEF, a message for registering capability as a VFL server for an Analytics ID; and transmit the request message for discovering at least one NWDAF.

[0139] In examples of the present disclosure, the AF entity may be caused to transmit, to the discovered NWDAF via the NEF, the VFL preparation request message, wherein the external NWDAF ID of the discovered NWDAF is mapped to an internal NWDAF ID of the discovered NWDAF by the NEF and a generic public subscription identifier (GPSI) of a terminal is mapped to a subscription permanent identifier (SUPI) of the terminal by the NEF; and receive, from the NWDAF via the NEF, a response message corresponding to the VFL preparation request message, wherein the internal NWDAF ID of the discovered NWDAF is mapped to the external NWDAF ID of the discovered NWDAF, by the NEF.

[0140]

[0141] Therefore, to reduce NEF workload and improve 5CG privacy, certain examples of the present disclosure allow the VFL NWDAF client to aggregate the VFL preparation responses from multiple other VFL clients to determine the intersection of sample IDs. Then the NEF will only map the internal IDs of the intersected Samples to external IDs, which reduces NEF exposure workload and lowers the risk of 5GC privacy leakage.

[0142]

[0143]

[0144] FIG. 1 illustrates an aspect of the background subject matter.

[0145] FIG. 2 illustrates an aspect of the background subject matter.

[0146] FIG. 3 illustrates an aspect of the background subject matter.

[0147] FIG. 4 illustrates an aspect of the background subject matter.

[0148] FIG. 5 illustrates a procedure for support of sample ID intersection for VFL according to an example of the present disclosure.

[0149] FIG. 6 illustrates a registration and discovery procedure for VFL when NWDAF or trusted AF is acting as VFL server according to an example of the present disclosure.

[0150] FIG. 7 illustrates registration and discovery procedure for VFL when an untrusted AF is acting as VFL server and NWDAF(s) are the VFL clients according to an example of the present disclosure.

[0151] FIG. 8 illustrates an example method of a VFL client (e.g. NWDAF) in a communication network.

[0152] FIG. 9 illustrates an example method of a first network entity (e.g. NWDAF) in a communication network.

[0153] FIG. 10 illustrates an example method of a VFL client (e.g. NWDAF) in a communication network.

[0154] FIG. 11 illustrates an example method of a first network entity (e.g. NEF) in a communication network.

[0155] FIG. 12 illustrates an example method of a first network entity (e.g. VFL client, VFL client NWDAF) in a communication network.

[0156] FIG. 13 illustrates an exemplary network entity that may be used in examples of the present disclosure.

[0157] Figure 14 (corresponding to Figure 6.2H.2.2-1 of TS　23.288) shows preparation procedure for Vertical Federated Learning when untrusted AF is the VFL Server.

[0158]

[0159] The following description of examples of the present disclosure, with reference to the accompanying drawings, is provided to assist in a comprehensive understanding of the present invention, as defined by the claims.　 The description includes various specific details to assist in that understanding but these are to be regarded as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the examples described herein can be made.

[0160] The same or similar components may be designated by the same or similar reference numerals, although they may be illustrated in different drawings.

[0161] Detailed descriptions of techniques, structures, constructions, functions or processes known in the art may be omitted for clarity and conciseness, and to avoid obscuring the subject matter of the present disclosure.

[0162] The terms and words used herein are not limited to the bibliographical or standard meanings, but, are merely used to enable a clear and consistent understanding of the examples disclosed herein.

[0163] Throughout the description and claims, the words "comprise", "contain" and "include", and variations thereof, for example "comprising", "containing" and "including", means "including but not limited to", and is not intended to (and does not) exclude other features, elements, components, integers, steps, processes, functions, characteristics, and the like.

[0164] Throughout the description and claims, the singular form, for example "a", "an" and "the", encompasses the plural unless the context otherwise requires. For example, reference to "an object" includes reference to one or more of such objects.

[0165] Throughout the description and claims, language in the general form of "X for Y" (where Y is some action, process, function, activity or step and X is some means for carrying out that action, process, function, activity or step) encompasses means X adapted, configured or arranged specifically, but not necessarily exclusively, to do Y.

[0166] Features, elements, components, integers, steps, processes, functions, characteristics, and the like, described in conjunction with a particular aspect, embodiment, example or claim are to be understood to be applicable to any other aspect, embodiment, example or claim disclosed herein unless incompatible therewith.

[0167] The following examples are applicable to, and use terminology associated with, 3GPP 5G.　However, the skilled person will appreciate that the techniques disclosed herein are not limited to these examples or to 3GPP 5G, and may be applied in any suitable system or standard, for example one or more existing and / or future generation wireless communication systems or standards. The skilled person will appreciate that the techniques disclosed herein may be applied in any existing or future releases of　3GPP 5G NR or any other relevant standard.

[0168] For example, the functionality of the various network entities and other features disclosed herein may be applied to corresponding or equivalent entities or features in other　communication systems or standards. Corresponding or equivalent entities or features may be regarded as entities or features that perform the same or similar role, function, operation or purpose within the network. For example, the functionality of a VFL server, VFL server AF, VFL client, or VFL client NWDAF in the examples below may be applied to any other suitable type of entity performing similar or equivalent functions.

[0169] The skilled person will appreciate that certain examples of the present disclosure may not be directly related to standardization but rather proprietary implementation of some of the VFL functions or non-VFL related functions of NR Rel-17 and beyond networks.

[0170] The skilled person will appreciate that the present invention is not limited to the specific examples disclosed herein. For example:

[0171] ● The techniques disclosed herein are not limited to 3GPP 5G.

[0172] ● The techniques disclosed herein are not limited to VFL.

[0173] ● One or more entities in the examples disclosed herein may be replaced with one or more alternative entities performing equivalent or corresponding functions, processes or operations.

[0174] ● One or more of the messages in the examples disclosed herein may be replaced with one or more alternative messages, signals or other type of information carriers that communicate equivalent or corresponding information.

[0175] ● One or more further elements, entities and / or messages may be added to the examples disclosed herein.

[0176] ● One or more non-essential elements, entities and / or messages may be omitted in certain examples.

[0177] ● The functions, processes or operations of a particular entity in one example may be divided between two or more separate entities in an alternative example.

[0178] ● The functions, processes or operations of two or more separate entities in one example may be performed by a single entity in an alternative example.

[0179] ● Information carried by a particular message in one example may be carried by two or more separate messages in an alternative example.

[0180] ● Information carried by two or more separate messages in one example may be carried by a single message in an alternative example.

[0181] ● The order in which operations are performed may be modified, if possible, in alternative examples.

[0182] ● The transmission of information between network entities is not limited to the specific form, type and / or order of messages described in relation to the examples disclosed herein.

[0183] The following exception has been agreed upon in SP-241513 / S2-2413037:

[0184] Vertical Federated Learning (VFL)

[0185] For the case where AF is untrusted whether additional functionality in NEF is needed during the VFL preparation

[0186] When untrusted AF is acting as the VFL server, as it has been documented in clause 6.2H.2.2.2 of TS 23.288 (see FIG. 4 above), the NEF will forward the VFL preparation response from every VFL client individually to the server by mapping a list of internal sample IDs to external IDs. Considering the potential high number of VFL clients and the sample IDs in each VFL preparation response, NEF will have very high workload on sample ID mapping and VFL preparation response exposure.

[0187] Furthermore, exposing the supported sample IDs of every single NWDAF VFL clients can potentially increase the risk of 5GC privacy leaking.

[0188] Therefore, to reduce NEF workload and improve 5CG privacy, certain examples of the present disclosure allow the VFL NWDAF client to aggregate the VFL preparation responses from multiple other VFL clients to determine the intersection of sample IDs. Then the NEF will only map the internal IDs of the intersected Samples to external IDs, which reduces NEF exposure workload and lowers the risk of 5GC privacy leakage.

[0189] Certain examples of the present disclosure use one or more NWDAF VFL clients to aggregate the supporting sample IDs of other one or more candidate NWDAF VFL clients. The NWDAF VFL client aggregator performs sample ID intersection based on the multiple responses received from the other candidate NWDAF VFL clients. The intersected sample IDs are the sample IDs that can be supported by all of the other candidate NWDAF VFL clients and the VFL client aggregator itself.

[0190] The proposal can not only reduce NEF workload on mapping internal NWDAF ID into external ID when sending VFL preparation response, but also lower the risk of 5GC privacy leakage by aggregating the supporting sample IDs of every single VFL client into one intersected sample ID list.

[0191] Terms

[0192] VFL process / VFL operationincludes one or more of the following procedures: VFL discovery and selection, VFL preparation, VFL training, VFL inference.

[0193] VFL client aggregator: a VFL client that aggregates the information from other VFL clients, and / or splits the request from VFL server to one or more indicated indirect VFL clients, e.g. the information received during VFL discovery and selection, VFL preparation, VFL training, VFL inference.

[0194] Indirect VFL client: the VFL clients who do not have direct contact with VFL server, and / or who contact the VFL server via VFL aggregator, e.g. during VFL discovery and selection, VFL preparation, VFL training, VFL inference.

[0195] Sample ID intersection:the sample IDs that can be supported by all of the candidate VFL clients (i.e. the client aggregator and each of the indirect VFLs). The sample IDs might be within the candidate sample IDs provided by the VFL server, or some of the sample IDs might be outside of the candidate sample IDs provided by the VFL server

[0196] Certain examples of the present disclosure may be provided in the form of an apparatus / device / network entity configured to perform one or more defined network functions and / or a method therefor.　Such an apparatus / device / network entity may comprise one or more elements, for example one or more of receivers, transmitters, transceivers, processors, controllers, modules, units, and the like, each element configured to perform one or more corresponding processes, operations and / or method steps for implementing the techniques described herein. For example, an operation / function of X may be performed by a module configured to perform X (or an X-module).　Certain examples of the present disclosure may be provided in the form of a system (e.g. a network) comprising one or more such apparatuses / devices / network entities, and / or a method therefor. For example, in the following examples, a network may include one or more VFL servers and one or more VFL clients.

[0197] It will be appreciated that examples of the present disclosure may be realized in the form of hardware, software or a combination of hardware and software.　Certain examples of the present disclosure may provide a computer program comprising instructions or code which, when executed, implement a method, system and / or apparatus in accordance with any aspect, claim, example and / or embodiment disclosed herein. Certain embodiments of the present disclosure provide a machine-readable storage storing such a program.

[0198] FIG. 5 illustrates a procedure for support of sample ID intersection for VFL according to an example of the present disclosure.

[0199] In the present disclosure, an untrusted AF VFL server is used as an example. But the VFL server can be also (trusted) AF and / or NWDAF.

[0200] 1-　　　 　　 　 The untrusted AF as VFL server sends VFL preparation requests to the candidate VFL client(s), using Nnef_VFLTrainingRequest_Request to check if the VFL client(s) can meet the ML Model training requirement (which includes the Analytics ID, list of sample IDs, and optionally Feature ID, etc.). The external NWDAF IDs obtained in the discovery procedure (see clause　6.2H.2.1.1 of TS　23.288 and FIG. 1) is included to indicate the target NWDAFs. The suggested VFL Interoperability Information to negotiate the intermediate results that will be used in training.

[0201] 1a-1c.　　　 　　 The untrusted AF as VFL server sends VFL preparation requests to each candidate VFL client(s) via NEF separately.

[0202] 1d.　　　 　　 　　 Alternatively, the untrusted AF as VFL server sends one VFL preparation request to NEF, the NEF sends one VFL preparation request to one or more selected NWDAF VFL client aggregator(s). The VFL server indicates the external NWDAF IDs of VFL client aggregator and the other of other corresponding candidate VFL client(s) to the NEF.

[0203] The untrusted AF as VFL server may group the candidate VFL clients, e.g. some VFL client will be contacted via another VFL client by the untrusted AF server (via NEF), or some VFL clients can be contacted directly by the untrusted AF server (via NEF).

[0204] When the VFL server is an untrusted AF, the AF invokes Nnef_VFLTraining_Request to the NEF. The VFL server includes a flag / indication to indicate that the NWDAF IDs are the IDs of NWDAF VFL client aggregator(s) and / or indirect VFL clients; or the server may include a flag / indication to indicate VFL client aggregator(s) and / or indirect VFL clients are needed / configured.

[0205] The VFL server may indicate one or more VFL client aggregators and their corresponding indirect VFL clients. E.g. VFL client aggregator 1 will contact to corresponding indirect VFL clients {1, ..., n}; VFL client aggregator 2 will contact to corresponding indirect VFL clients {a, ... ,z};

[0206] The untrusted AF VFL server may choose the VFL client aggregator(s) and / or indirect VFL clients based on at least one of the following:

[0207] preconfigured information, e.g. by the operator, service provider; e.g. the operator may preconfigure one or more NWDAF VFL clients that can act as VFL aggregators; based on this information, the VFL server can choose to select one or more VFL clients as VFL aggregators during VFL preparation stage.

[0208] Prior knowledge, e.g. the stored information collected from one or more previous VFL processes;

[0209] the VFL client capability information, e.g. the information registered in NRF during NF registration (e.g. whether the VFL client can act as aggregator(s) and / or indirect VFL clients or not), which is requested by VFL server during VFL client discovery stage, etc.

[0210] New capability information is needed: for example VFL client type of VFL client aggregator(s) and / or indirect VFL clients.

[0211] From the perspective of VFL server, the server is required to choose VFL client aggregator(s) and / or indirect VFL clients. Therefore, new capability of VFL server is also required.

[0212] All the new capability information might be registered to NRF during VFL registration procedure, as detailed by 6.2H.2.1.1　of TS　23.288　(Registration and Discovery procedure for Vertical Federated Learning when NWDAF or trusted AF is acting as the VFL server, see FIG. 1)and6.2H.2.1.2　of TS　23.288　(Registration and Discovery procedure for Vertical Federated Learning when untrusted AF is acting as the VFL server, see FIG. 2)in the present disclosure. Therefore, the above new capability of VFL client aggregator(s), indirect VFL clients, and VFL severs can be required during VFL discovery.

[0213] Based on the new capability information of VFL client aggregator(s) and / or indirect VFL clients received by VFL server during discovery, the VFL server is able to choose VFL client aggregator(s) and / or indirect VFL clients for the corresponding procedures; e.g. for sample ID intersection, for aggregating intermediates training and / or inference results of VFL clients, for sharing intermediates training and / or inference results with other VFL clients, etc.

[0214] NOTE　: subject to operators' policies, the untrusted AF maybe required to contact the other candidate NWDAF VFL client(s) via a NWDAF VFL client aggregator to protect 5GC privacy.

[0215] 2-　　 The NEF maps the external NWDAF and GPSI(s) to the internal NWDAF and SUPI(s).

[0216] 2a-2c.　　　 　　 The NEF sends VFL preparation request to the corresponding candidate internal NWDAF ID using Nnwdaf_VFLTraining_Request service with the same information as provided in step　1 in clause　6.2H.2.2.1 of TS　23.288 (see FIG. 3).

[0217] 2d-2e.　　　 　　 Alternatively, the NEF sends VFL preparation request to one or more VFL client aggregator. Then each VFL client aggregator splits the request and sends the VFL preparation request to the other corresponding candidate NWDAF VFL client(s).

[0218] The other corresponding candidate NWDAF VFL client(s) might be the indirect VFL clients.

[0219] The VFL client aggregator sends the VFL preparation request based on the NWDAF IDs included in the NEF request, e.g. Nnef_VFLTraining_Request. The NEF may include flags / indications / information to inform the VFL client aggregator that it is selected as VFL client aggregator, and / or needs to split / forward the VFL preparation request to the indicated candidate indirect VFL clients. Splitting / the VFL preparation request may sending each indicated candidate VFL client the VFL preparation request with the IDs of the other indicated candidate VFL clients removed. In detail, when the VFL client aggregator splits and / or forwards the VFL preparation request to the other candidate VFL client, it will remove the list of NWDAF IDs received from the NEF and sends VFL preparation request to the NWDAFs as indicated by the list of NWDAF IDs.

[0220] If one or more (list of) NWDAF IDs is included in the NEF request, the VFL client aggregator may understand it is a VFL client aggregator and will need to split request to other VFL clients and aggregate response from them.

[0221] Or explicit flags / indications may be used to inform the VFL client aggregator.

[0222] 3-　　 Same as step　2 in clause　6.2H.2.2-1 of TS　23.288 (see FIG. 3). Each VFL Client checks if it can meet the ML Model training requirement. Each VFL Client ID checks the list of sample IDs suggested by the VFL Server and then provides to the VFL Server the list of sample IDs that it can accept out of the sample IDs and required criteria for sample alignment suggested by the VFL Server and satisfying the required criteria for sample alignment. Each VFL Client checks the VFL Interoperability Information and determines which VFL Interoperability information that the VFL Client supports. The VFL Clients provides the list of supported Feature IDs, which is associated to the VFL Interoperability information to the VFL Server, if available, or VFL Server may know the supported FeatureIDs for a VFL Client based on configuration.

[0223] In the present disclosure, both the　indirect VFL clients and VFL aggregator may need to check whether they can meet the VFL requirement or not.

[0224] 4-　　 The candidate VFL clients invokes Nnwdaf_VFLTraining_Response to indicate whether it accepts the ML Model training requirements, the VFL Client can also indicate that it cannot join the FL process.

[0225] 4a-4c. Same as step　3 in clause　6.2H.2.2-1 of TS　23.288 (see FIG. 3), if the VFL server sends preparation request to each candidate VFL client separately in step　1.

[0226] 4d-4f.　　　 　　 Alternatively, the candidate VFL clients in step　2e send responses to the candidate VFL client aggregator. The VFL client aggregator aggregates the responses from multiple candidate NWDAF VFL clients and determines the sample ID intersection among the multiple NWDAF VFL clients and itself. For example, to determine the sample ID intersection, the VFL client aggregator determines which sample IDs are indicated as supported in every response from the candidate VFL clients, and determines if any of these fully supported sample IDs are supported by itself. Then the VFL client aggregator sends the intersected sample IDs and the other information in step 3 of clause 6.2H.2.2-1 of TS　23.288 (see FIG. 3) to the VFL server via NEF.

[0227] ● Each VFL client aggregator may only send one list of intersected sample IDs to the VFL server via NEF, rather than the lists of supporting sample IDs of each individual candidate VFL client.

[0228] ● Only reporting the intersected sample IDs to the untrusted AF server can significantly lower the network privacy leakage, as the server will not know the exact sample IDs that can be supported by each single VFL client; and therefore, it can avoid the untrusted AF to derive the network topology and other sensitive information related to deployment.

[0229] ● The VFL aggregator may also include the NWDAF IDs of each indirect VFL clients to the response to NEF. In the response, the VFL aggregate may associated the response from each indirect VFL clients to their NWDAF IDs.

[0230] 5-　　 The NEF maps the internal NWDAF and SUPI(s) to the external NWDAF and GPSI(s). The NEF sends the Nnef_VFLTraining_Request Response to the VFL Server with the same information as provided in step　4.

[0231] If VFL aggregator is selected and used, in step 4, the aggregate may only report one list of the intersected sample IDs supported by the VFL aggregator and the corresponding indirect VFL clients. Compared to sending request to each VFL client separately, using VFL client aggregator can reduce the NEF workload on translating internal sample ID to external sample ID significantly. In this case, the NEF does not need to translate the supporting sample IDs reported by every single VFL clients.

[0232] 6-　　 Same as step　4 in clause　6.2H.2.2-1 of TS　23.288 (see FIG. 3). The VFL server determines the final list of samples considering the samples that all selected VFL clients support, if used the Feature ID per VFL client and VFL Interoperability Information to be used for training and provide them to the selected VFL Clients at the start of the training phase, as described in clause 6.2H.2.3. of TS　23.288.

[0233]

[0234] VFL client aggregator and indirect VFL client selection

[0235] Registration and Discovery procedure for Vertical Federated Learning when NWDAF or trusted AF is acting as the VFL server

[0236] FIG. 6 illustrates a registration and discovery procedure for VFL when NWDAF or trusted AF is acting as VFL server (e.g. of TS　23.288) according to an example of the present disclosure.

[0237] Steps 1 to 3 are the NWDAF and AF Registration procedures when the VFL server is NWDAF or a trusted AF.

[0238] 1a.　VFL Server NWDAF / trusted AF registers to NRF with its NF profile, which includes NF Type (i.e. NWDAF type or AF type), Analytics ID(s), service area if available, VFL capability information per analytics ID and VFL interoperability indicator(s) and optional supported feature ID(s), and other parameters as described in clause　5.2 of TS　23.288.

[0239] In an example of the present disclosure, the VFL Server NWDAF / trusted AF may also register whether they have the capability of select VFL client aggregator and / or indirect VFL client, e.g. as part of the NF Type: can support NWDAF / AF server and optionally selection of VFL client aggregator and / or indirect VFL client.

[0240] 1b.　NWDAF as VFL client registers to NRF with its NF profile, which includes NF Type (i.e. NWDAF type), Analytics ID(s), service area if available, VFL capability information per analytics ID and VFL interoperability indicator(s) and optional supported feature ID(s), as defined in clause　5.2 of TS　23.288.

[0241] In an example of the present disclosure, the VFL NWDAF client may also register whether they have the capability of acting as VFL client aggregator and / or indirect VFL client, e.g. as part of the NF Type.

[0242] 2.　　 The NRF receives the registrations from VFL server and VFL client(s), and stores their NF profile.

[0243] 3.　　 The NRF sends registration response to VFL server and VFL client(s).

[0244] Steps 4 to 6 are the NWDAF, AF and NEF Discovery procedures when the VFL server is NWDAF or trusted AF.

[0245] 4-6.　　 The VFL server determines that the ML Model requires VFL based on e.g. operator policy, Analytics ID, VFL interoperability　indicator(s)　and Service Area.

[0246] Once the VFL Server is determined, the VFL Server discovers other NWDAF(s) and / or AF(s) as VFL Client from NRF by invoking the Nnrf_NFDiscovery_Request service operation. The following criteria might be used: NF type(s) (i.e. NWDAF type, AF type, or NEF type), Analytics ID, VFL capability type (i.e.VFL client), VFL interoperability indicator(s), Time Period of Interest, optional feature ID(s) and optional Service Area. The AF(s) discovered by the VFL Server may be trusted and / or untrusted, the NF type may be AF type when the AF(s) as VFL Client are trusted, and the NF type may be NEF type when the AF(s) as VFL Client are untrusted, and the NF type may contain both AF type and NEF type if both trusted AF and untrusted AF are involved as VFL clients.

[0247] In an example of the present disclosure, the criteria may also include whether VFL client can be VFL client aggregator and / or indirect VFL client, e.g. for NWDAF VFL clients.

[0248] Registration and Discovery procedure for Vertical Federated Learning when untrusted AF is acting as the VFL server

[0249] FIG. 7 illustrates registration and discovery procedure for VFL when an untrusted AF is acting as VFL server and NWDAF(s) are the VFL clients (e.g. of TS　23.288) according to an example of the present disclosure.

[0250] Steps 1 to 3 are the NWDAF and AF Registration procedures when the VFL server is untrusted AF.

[0251] 1.　　 Same as the step　1b,in Registration and Discovery procedure for Vertical Federated Learning when NWDAF or trusted AF is acting as the VFL server(see FIG. 6), NWDAF as VFL client registers to NRF with its NF profile, it may include those analytics IDs on which it supports to do VFL with AF as VFL server.

[0252] 2-3.　　 Same as the steps 2-3,in　Registration and Discovery procedure for Vertical Federated Learning when NWDAF or trusted AF is acting as the VFL server(see FIG. 6). Based on operator local policies, the AF may register its capability as VFL server for an Analytics ID to NEF via OAM configuration, this is used to make it possible for NWDAF to trigger the AF to train a model. The AF may also register an AnalyticsID in NRF e.g. when a trained model is available.

[0253] In an example of the present disclosure, the VFL Server untrusted AF may also register whether they have the capability of select VFL client aggregator and / or indirect VFL client, e.g. as part of the NF Type: can support AF server and optionally selection of VFL client aggregator and / or indirect VFL client (e.g. from NWDAF VFL clients).

[0254] Steps 4-10 are the NWDAF Discovery procedures when the VFL server is an untrusted AF.

[0255] 4.　　 Untrusted AF acting as the VFL server determines that VFL operations are required and the NWDAF(s) as VFL client(s) are required. The AF sends a Nnef_NFDiscovery_Request for the VFL client(s) to the NEF and for each client provides selection criteria: Analytics ID, required NF type (i.e. NWDAF type), VFL capability type (i.e. VFL client), VFL interoperability indicator(s), Time Period of Interest, optional required feature ID(s), and optional Service Area.

[0256] In an example of the present disclosure, untrusted AF acting as the VFL server may determine that VFL operations are required and the NWDAF(s) as VFL client aggregator and / or indirect VFL clients are required. The AF sends a Nnef_NFDiscovery_Request to the determined VFL client aggregator for response of the VFL client aggregator and other indirect VFL clients to the NEF. The untrusted AF server may also provide the same or different selection criteria to VFL client aggregator and indirect VFL clients. Other than the following information: Analytics ID, required NF type (i.e. NWDAF type), VFL capability type (i.e. VFL client), VFL interoperability indicator(s), Time Period of Interest, optional required feature ID(s), and optional Service Area; the VFL server may also provide NF / NWDAF type is NWDAF VFL client aggregator and / or other indirect VFL clients.

[0257] 5.　　 The NEF checks based on configured policies whether the AF is entitled to request single or multiple VFL client(s)　 for the Analytics ID.

[0258] 6-7.　　 The NEF discovers VFL client(s) (i.e. NWDAF(s)) on behalf of the AF from the NRF by invoking the Nnrf_NFDiscovery_Request using the selection criteria provided by the AF as defined in step　4.

[0259] 8.　　 The NEF selects NWDAF(s) , e.g., by using the NWDAF(s) NF profile and parameters including load / priority / capacity that are capable of acting as VFL client(s) and matching the received selection criteria such as location information. The NEF anonymizes NWDAF instances ID(s) and assigns external NWDAF ID(s) for each selected NWDAF intance as VFL client. The NEF stores the external NWDAF ID(s) together with information how to reach the NWDAFs.

[0260] 9.　　 The NEF sends Nnef_NFDiscovery_Request response only including the external NWDAF ID(s) as selected in step 8 and information how they relate to corresponding selection criteria provided by the AF, VFL interoperability indicator(s),　Time Period of Interest, optional feature ID(s) and optional Service Area to the untrusted AF.

[0261] 10.　The AF stores the received external NWDAF ID(s) and uses it subsequent interactions with the NEF to indicate the target VFL client.

[0262] Service operation to support sample ID intersection

[0263] Y.2.5　　　 Nnef_VFLTraining_Request service operation

[0264] Service operation name:　Nnef_VFLTraining_Request

[0265] Description:In preparation of VFL training, requests NEF to check at untrusted AF acting as VFL client if it can support requirements for VFL.

[0266] Inputs, Required:

[0267] Analytics ID;

[0268] For NWDAF as VFL client, external NWDAF ID

[0269] Inputs, Optional:

[0270] ●NWDAF client aggregator NWDAF IDs and / or the associated VFL aggregator flag to indicate the IDs are for client aggregator.

[0271] ●one or more (list of) external NWDAF IDs for indirect NWDAF VFL clients, the associated NWDAF client aggregator, indirect VFL client flag to indicate the IDs are for indirect VFL clients.

[0272] Outputs Required:When the request is accepted:　list of sample IDs, VFL Interoperability Information. When the request is not accepted, an error response with cause code (e.g. NWDAF does not meet the VFL training requirements

[0273] The output may also include　 NWDAF IDs of indirect VFL clients and one list of the intersected sample IDs (external IDs) when NWDAF client aggregator and indirect NWDAF VFL clients are selected and configured.

[0274] Outputs, Optional:　none

[0275] 7.X　　　 　 Nnwdaf_VFLTraining Service

[0276] 7.X.1　　　 　　 General

[0277] Service Description: This service is provided by an NWDAF acting as VFL client and enables an NWDAF VFL server, an AF VFL server or an NEF acting on its behalf as consumer to request the NWDAF to participate in VFL preparation and model training as VFL client.

[0278] If VFL client aggregator and / d or indirect VFL clients are selected and configured, the Nnwdaf_VFLTraining can be also used by the NWDAF VFL aggregator to request the NWDAF to participate in VFL preparation and model training as indirect VFL client.

[0279] 7.X.5　　　 　　 Nnwdaf_VFLTraining_Request service operation

[0280] Service operation name:　Nnwdaf_VFLTraining_Request

[0281] Description:In preparation of VFL training, requests NWDAF VFL(indirect)client to check if it can support requirements for VFL.

[0282] Inputs, Required:VFL operation criteria detailed in previous clauses.

[0283] Inputs, Optional:　none

[0284] Outputs Required:When the request is accepted:the supporting sample IDs and / or feature IDs of the VFL client.

[0285] When the request is not accepted, an error response with cause code

[0286] Outputs, Optional:　none

[0287] FIG. 8 illustrates an example method of a VFL client (e.g. NWDAF) in a communication network.

[0288] In step 802, the method comprises registering to a first network entity (e.g. NRF) based on profile information (e.g. NF profile) of the VFL client.

[0289] The profile information includes capability information indicating that the VFL client is capable of aggregating intermediate results of other VFL clients.

[0290] FIG. 9 illustrates an example method of a first network entity (e.g. NWDAF) in a communication network.

[0291] In step 902, the method comprises storing profile information (e.g. NF profile) of a VFL client (e.g. NWDAF).

[0292] The profile information includes capability information indicating that the VFL client is capable of aggregating intermediate results of other VFL clients.

[0293] FIG. 10 illustrates an example method of a VFL client (e.g. NWDAF) in a communication network.

[0294] In step 1002, the method comprises transmitting, to one or more other VFL clients, VFL preparation requests.

[0295] In step 1004, the method comprises receiving, from the one or more other VFL clients, responses to the VFL preparation requests.

[0296] In step 1006, the method comprises aggregating the responses from the one or more other VFL clients.

[0297] FIG. 11 illustrates an example method of a first network entity (e.g. NEF) in a communication network.

[0298] In step 1102, the method comprises receiving, from a VFL server (e.g. AF), a VFL request, wherein the VFL request includes identification information identifying one or more target VFL clients.

[0299] In step 1104, the method comprises transmitting, to at least one of the one or more target VFL clients based on the identification information, a VFL preparation request.

[0300] FIG. 12 illustrates an example method of a first network entity (e.g. VFL client, VFL client NWDAF) in a communication network.

[0301] In step 1202, the method comprises receiving, from a second network entity (e.g. VFL server, VFL server AF), a preparation request comprising information indicating a plurality of indirect VFL clients.

[0302] In step 1204, the method comprises sending, to each of the indicated plurality of indirect VFL clients, a split or forwarded preparation request.

[0303] In step 1206, the method comprises receiving, from the indicated plurality of indirect VFL clients, a preparation request response including a list of supported sample IDs.

[0304] In step 1208, the method comprises determining a sample ID intersection based on the received reparation request responses and the sample IDs supported by the first network entity.

[0305] In step 1210, the method comprises sending, to the second network entity, the sample ID intersection.

[0306] FIG. 13 is a block diagram of an exemplary network entity (e.g. VFL server, VFL server AF, NEF, VFL client, VFL client NWDAF) that may be used in examples of the present disclosure. The skilled person will appreciate that the network entity illustrated in FIG. 13 may be implemented, for example, as a network element on a dedicated hardware, as a software instance running on a dedicated hardware, or as a virtualised function instantiated on an appropriate platform, e.g. on a cloud infrastructure.

[0307] The network entity 1300 comprises a processor 1301 (or controller), a transmitter 1303 and a receiver 1305. The receiver 1305 is configured for receiving one or more messages from one or more other network entities. The transmitter 1303 is configured for transmitting one or more messages to one or more other network entities. The processor 1301 is configured for performing operations as described above.

[0308] In an example of the present disclosure, there is provided a method of a first network entity (e.g. VFL client, VFL client NWDAF) in a communication network, the method comprising: receiving, from a second network entity (e.g. VFL server, VFL server AF), a preparation request comprising information indicating a plurality of indirect VFL clients; sending, to each of the indicated plurality of indirect VFL clients, a split or forwarded preparation request; receiving, from the indicated plurality of indirect VFL clients, a preparation request response including a list of supported sample IDs; determining a sample ID intersection based on the received reparation request responses and the sample IDs supported by the first network entity; sending, to the second network entity, the sample ID intersection.

[0309] In a first example, there is provided a method of a VFL client (e.g. NWDAF) in a communication network, the method comprising: registering to a first network entity (e.g. NRF) based on profile information (e.g. NF profile) of the VFL client; wherein the profile information includes capability information indicating that the VFL client is capable of aggregating intermediate results of other VFL clients.

[0310] In a second example, there is provided a method of a first network entity (e.g. NRF) in a communication network, the method comprising: storing profile information (e.g. NF profile) of a VFL client (e.g. NWDAF); wherein the profile information includes capability information indicating that the VFL client is capable of aggregating intermediate results of other VFL clients.

[0311] In a third example, there is provided the method of the first or second example, wherein the profile information includes analytics IDs on which the VFL client supports VFL operation with a VFL server (e.g. AF).

[0312] In a fourth example, there is provided the method of the third example, wherein the VFL server is an untrusted AF.

[0313] In a fifth example, there is provided a method of a VFL client (e.g. NWDAF) in a communication network, the method comprising: transmitting, to one or more other VFL clients, VFL preparation requests; receiving, from the one or more other VFL clients, responses to the VFL preparation requests; and aggregating the responses from the one or more other VFL clients.

[0314] In a sixth example, there is provided the method of the fifth example, further comprising: receiving, from a first network entity (e.g. NEF), a VFL preparation request; and in response to receiving the VFL preparation request, determining to transmit, to the one or more other VFL clients, the VFL preparation requests.

[0315] In a seventh example, there is provided the method of the fifth or sixth example, wherein aggregating the responses from the one or more other VFL clients comprises aggregating the responses from the one or more other VFL clients and information of the VFL client.

[0316] In an eighth example, there is provided the method of any one of the fifth to seventh examples, wherein aggregating the responses from the other or more other VFL clients comprises aggregating intermediate results included in the responses from the one or more other VFL clients.

[0317] In a ninth example, there is provided the method of any one of the fifth to eighth examples, further comprising: transmitting, to a VFL server (e.g. AF), a VFL response based on aggregating the responses from the one or more other VFL clients.

[0318] In a tenth example, there is provided the method of the ninth example, wherein transmitting, to the VFL server, a VFL response based on aggregating the responses from the one or more other VFL clients comprises transmitting, to the VFL server, a VFL response based on aggregating intermediate results included in the responses from the one or more other VFL clients.

[0319] In an eleventh example, there is provided the method of any one of the fifth to tenth examples, wherein transmitting, to the one or more other VFL clients, the VFL preparation requests, comprises transmitting, to each of the one or more other VFL clients, a split or forwarded VFL preparation request.

[0320] In a twelfth example, there is provided the method of any one of the sixth to eleventh examples, wherein the VFL preparation request received from the first network entity (e.g. from a VFL server via the first network entity) comprises information indicating the one or more other VFL clients.

[0321] In a thirteenth example, there is provided the method of any one of the fifth to twelfth examples, wherein the responses received from the one or more other VFL clients each include a list of supported sample IDs; wherein aggregating the responses comprises determining a sample ID intersection based on the responses and the sample IDs supported by the first network entity; and wherein the method further comprises transmitting, to a VFL server, the sample ID intersection.

[0322] In a fourteenth example, there is provided the method of the tenth example, wherein the VFL server is an untrusted AF.

[0323] In a fifteenth example, there is provided the method of any one of the first to fourteenth examples, wherein the other VFL clients are NWDAFs.

[0324] In a sixteenth example, there is provided a method of a first network entity (e.g. NEF) in a communication network, the method comprising: receiving, from a VFL server (e.g. AF), a VFL request, wherein the VFL request includes identification information identifying one or more target VFL clients; and transmitting, to at least one of the one or more target VFL clients based on the identification information, a VFL preparation request.

[0325] In a seventeenth example, there is provided the method of the sixteenth example, wherein transmitting, to the at least one of the one or more target VFL clients based on the identification information, the VFL preparation request comprises: mapping external NWDAF and GPSI(s) included in the identification information to internal NWDAF and SUPI(s); and transmitting the VFL preparation request to a corresponding internal NWDAF ID.

[0326] In an eighteenth example, there is provided the method of the seventeenth example, further comprising: receiving, from the at least one target VFL client, an aggregated response based on aggregating the responses from the one or more target VFL clients; and transmitting, to the VFL server, a VFL response based on the aggregated response.

[0327] In a nineteenth example, there is provided the method of the eighteenth example, wherein transmitting, to the VFL server, the VFL response based on the aggregated response comprises: mapping internal NWDAF and SUPI(s) to external NWDAF and GPSI(s); and transmitting to the VFL server, the VFL response based on the mapping.

[0328] In a twentieth example, there is provided the method of the sixteen example, further comprising: receiving, from the at least one target VFL client, a response including a sample ID intersection indicating sample IDs supported by the one or more target VFL clients; and transmitting, to the VFL server, a VFL response including the sample ID intersection.

[0329] In a twenty-first example, there is provided the method of any of the sixteenth to twentieth examples, wherein the target VFL clients are NWDAFs.

[0330] In a twenty-second example, there is provided a method of a first network entity (e.g. VFL client, VFL client NWDAF) in a communication network, the method comprising: receiving, from a second network entity (e.g. VFL server, VFL server AF), a preparation request comprising information indicating a plurality of indirect VFL clients; sending, to each of the indicated plurality of indirect VFL clients, a split or forwarded preparation request; receiving, from the indicated plurality of indirect VFL clients, a preparation request response including a list of supported sample IDs; determining a sample ID intersection based on the received reparation request responses and the sample IDs supported by the first network entity; and sending, to the second network entity, the sample ID intersection.

[0331] In a twenty-third example, there is provided a VFL client (e.g. NWDAF) configured to operate according to a method of any one of the first, third to fifteenth, or twenty-second examples.

[0332] In a twenty-fourth example, there is provided a first network entity (e.g. NRF, NEF) configured to cooperate with a first network entity of the twenty-third example according to a method of any one of the second to fourth or sixteenth to twenty-first examples.

[0333] In a twenty-fifth example, there is provided a network or wireless communication system comprising a first network entity according to the twenty-third example and a second network entity according to the twenty-fourth example.

[0334] In a twenty-sixth example, there is provided a computer program comprising instructions which, when the program　 is executed by a computer or processor, cause the computer or processor to carry out a method according to any one of the first to twenty-second examples.

[0335] In a twenty-seventh example, there is provided a computer or processor-readable data carrier having stored thereon a computer program according to　the twenty-sixth example.

[0336] While the invention has been shown and described with reference to certain examples, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the scope of the invention, as defined by the appended claims.

[0337] Certain examples of the present disclosure provide one or more techniques as disclosed in the appended annex to the description. The skilled person will appreciate that any of these techniques may be applied in combination with any of the techniques described above and illustrated in the Figures.

[0338] By way of further explanation, examples of how the existing 3GPP standards might be modified in view of certain aspects of the present disclosure are provided.

[0339] Hereinafter, the reasons for the modifications will be described.

[0340] The following exception has been agreed upon in SP-241513 / S2-2413037:

[0341] Vertical Federated Learning (VFL)

[0342] For the case where AF is untrusted whether additional functionality in NEF is needed during the VFL preparation

[0343] When untrusted AF is acting as the VFL server, as it has been documented in clause 6.2H.2.2.2 of TS　23.288, the NEF will forward the VFL preparation response from every VFL client individually to the server by mapping a list of internal sample IDs to external IDs. Exposing the supported sample IDs of every single NWDAF VFL clients can potentially increase the risk of 5GC privacy leakage.

[0344] CR S2-2500432 was submitted to SA2#166AHE with proposing to enhance NEF to perform sample ID intersection. But the CR was noted due to concerns on NEF complexity.

[0345] In the present disclosure, modifications are proposed to use NWDAF VFL client to perform sample ID intersection, which will not add additional complexity of NEF but can still protect 5GC privacy.

[0346] Hereinafter, a summary of the proposed modifications will be provided.

[0347] 1. Remove EN on NEF work during VFL preparation in clause 6.2H.2.2.2 of TS　23.288.

[0348] 2. Introduce new steps into clause 6.2H.2.2.2 of TS　23.288 to support NWDAF VFL client to determine the sample ID intersection.

[0349] Hereinafter, the proposed modifications to clause 6.2H.2.2.2 (Preparation procedure for Vertical Federated Learning when untrusted AF is the VFL server) of TS　23.288 will be provided.

[0350] This clause specifies the preparation (including sample alignment) procedure for untrusted AF-initiated VFL scenarios between AF and NWDAF(s) within a single PLMN.

[0351] Figure 14 (corresponding to Figure 6.2H.2.2-1 of TS　23.288) shows preparation procedure for Vertical Federated Learning when untrusted AF is the VFL Server

[0352] 0. If UEs are to be used as samples, an untrusted AF may use the Nnef_MemberSelection service to ask the NEF to reduce the list of candidate UEs based on filter criteria it provides.

[0353] 1- The untrusted AF as VFL server sends VFL preparation requests to the candidate VFL client(s), using Nnef_VFLTrainingRequest_Request to check if the VFL client(s) can meet the ML Model training requirement (which includes the Analytics ID, list of sample IDs, and optionally Feature ID, etc.). The external NWDAF IDs obtained in the discovery procedure (see clause　6.2H.2.1.1) is included to indicate the target NWDAFs. The suggested VFL Interoperability Information to negotiate the intermediate results that will be used in training.

[0354] 1a-1c. The untrusted AF as VFL server sends VFL preparation requests to each candidate VFL client(s) via NEF separately.

[0355] 1d. Alternatively, the untrusted AF as VFL server sends one VFL preparation request to NEF by including the candidate VFL client NWDAF IDs. The candidate VFL client NWDAF IDs include the Ids of the VFL client aggregator and other VFL clients.

[0356] NOTE　: subject to operators' policies, the untrusted AF maybe required to contact the other candidate NWDAF VFL client(s) via one NWDAF VFL client to protect 5GC privacy. The VFL client aggregator might be pre-configured or selected by VFL server based on its stored information.

[0357] 2- The NEF maps the external NWDAF and GPSI(s) to the internal NWDAF and SUPI(s).

[0358] 2a-2c. The NEF sends VFL preparation request to the corresponding candidate internal NWDAF ID using Nnwdaf_VFLTraining_Request service with the same information as provided in step　1 in clause　6.2H.2.2.1 of TS　23.288.

[0359] 2d-2e. Alternatively, the NEF sends a VFL preparation request to one VFL client. Then this VFL client sends multiple VFL preparation requests to the other corresponding candidate NWDAF VFL client(s).

[0360] 3- Same as step　2 in clause　6.2H.2.2-1 of TS　23.288.

[0361] 4- The candidate VFL clients invokes Nnwdaf_VFLTraining_Response to indicate whether it accepts the ML Model training requirements, the VFL Client can also indicate that it cannot join the FL process.

[0362] 4a-4c. Same as step　3 in clause　6.2H.2.2-1 of TS　23.288, if the VFL server sends preparation request to each candidate VFL client separately in step　1.

[0363] 4d-4f. Alternatively, the candidate VFL clients in step　2e sends response to the VFL client aggregator. The VFL client aggregator aggregates the responses from multiple candidate NWDAF VFL clients and determines the sample ID intersection among the other candidate NWDAF VFL clients and itself. Then the VFL client aggregator sends the intersected sample IDs and the other information in step 3 of clause 6.2H.2.2-1 of TS　23.288 to the VFL server via NEF.

[0364] 5- The NEF maps the internal NWDAF and SUPI(s) to the external NWDAF and GPSI(s). The NEF sends the Nnef_VFLTraining_Request Response to the VFL Server with the same information as provided in step　4.

[0365] 6- Same as step　4 in clause　6.2H.2.2-1 of TS　23.288.

[0366] Abbreviations / definitions

[0367] In the present disclosure, the following abbreviations and definitions may be used.

[0368] 3GPP 3rd Generation Partnership Project

[0369] 5G 5th Generation

[0370] 5GC 5G Core

[0371] AF Application Function

[0372] AI Artificial Intelligence

[0373] AIML Artificial Intelligence and Machine Learning

[0374] AS Application Server

[0375] CN Core Network

[0376] CR Change Request

[0377] FL Federated Learning

[0378] GPSI Generic Public Subscription Identifier

[0379] ID Identity / Identifier

[0380] IP Internet Protocol

[0381] ML Machine Learning

[0382] NEF Network Exposure Function

[0383] NF Network Function

[0384] NR New Radio

[0385] NRF Network Repository Function

[0386] NG-RAN Next Generation Radio Access Network

[0387] NW Network

[0388] NWDAF Network Data Analytics Function

[0389] OAM Operations, Administration and Maintenance

[0390] PLMN Public Land Mobile Network

[0391] RAN Radio Access Network

[0392] SUPI Subscription Permanent Identifier

[0393] TS Technical Specification

[0394] UE User Equipment

[0395] VFL Vertical Federated Learn

Claims

1.A method performed by a network data analytics function (NWDAF) performing vertical federated learning (VFL) in a wireless communication system, the method comprising:transmitting, to a network repository function (NRF), a registration message including a network function (NF) profile of the NWDAF, wherein a capability of aggregating at least one intermediate result for at least one VFL of other NWDAF is included in the NF profile;receiving, from the NRF, a registration response corresponding to the registration message;receiving, from an application function (AF) via a network exposure function (NEF), a VFL preparation request message;transmitting, to at least one other NWDAF, at least one VFL preparation request message;receiving, from the at least one other NWDAF, at least one response, based on the at least one VFL preparation request message;aggregating the at least one received response; andtransmitting, to the AF via the NEF, the aggregated response.2.The method of claim 1,wherein an identity (ID) of the NWDAF selected as a VFL client is anonymized by the NEF and an external NWDAF ID of the NWDAF selected as a VFL client is assigned by the NEF; andwherein the external NWDAF ID of the NWDAF is stored, by the NEF, together with information associated with reaching the NWDAF.3.The method of claim 2, further comprising;receiving the VFL preparation request message, wherein the external NWDAF ID is mapped to an internal NWDAF ID of the NWDAF by the NEF and a generic public subscription identifier (GPSI) of a terminal is mapped to a subscription permanent identifier (SUPI) of the terminal by the NEF, based on a reception of the VFL preparation request message; andtransmitting the aggregated response, wherein the internal NWDAF ID of the NWDAF is mapped to the external NWDAF ID of the NWDAF, by the NEF, based on a transmission of the aggregated response.4.The method of claim 1,wherein a request for discovering at least one NWDAF is transmitted, from the AF to the NEF;wherein the at least one NWDAF is selected, by the NEF, according to at least one criterion for selecting at least one NWDAF; andwherein the at least one criterion is provided, to the NEF by the AF and is associated with the capability of aggregating.5.A network data analytics function (NWDAF) entity in a wireless communication system, the NWDAF entity comprising:at least one transceiver;at least one processor communicatively coupled to the at least one transceiver; andat least one memory, communicatively coupled to the at least one processor, storing instructions executable by the at least one processor individually or in any combination to cause the NWDAF entity to:transmit, to a network repository function (NRF), a registration message including a network function (NF) profile of the NWDAF, wherein a capability of aggregating at least one intermediate result for at least one VFL of other NWDAF is included in the NF profile;receive, from the NRF, a registration response corresponding to the registration message;receive, from an application function (AF) via a network exposure function (NEF), a VFL preparation request message;transmit, to at least one other NWDAF, at least one VFL preparation request message;receive, from the at least one other NWDAF, at least one response, based on the at least one VFL preparation request message;aggregate the at least one received response; andtransmit, to the AF via the NEF, the aggregated response.6.The NWDAF entity of claim 5;wherein an identity (ID) of the NWDAF selected as a VFL client is anonymized by the NEF and an external NWDAF ID of the NWDAF selected as a VFL client is assigned by the NEF; andwherein the external NWDAF ID of the NWDAF is stored, by the NEF, together with information associated with reaching the NWDAF.7.The NWDAF entity of claim 6, wherein the NWDAF entity is caused to;receive the VFL preparation request message, wherein the external NWDAF ID is mapped to an internal NWDAF ID of the NWDAF by the NEF and a generic public subscription identifier (GPSI) of a terminal is mapped to a subscription permanent identifier(SUPI) of the terminal by the NEF, based on a reception of the VFL preparation request message; andtransmit the aggregated response, wherein the internal NWDAF ID of the NWDAF is mapped to the external NWDAF ID of the NWDAF, by the NEF, based on a transmission of the aggregated response.8.The NWDAF entity of claim 5,wherein a request for discovering at least one NWDAF is transmitted, from the AF to the NEF;wherein the at least one NWDAF is selected, by the NEF, according to at least one criterion for selecting at least one NWDAF; andwherein the at least one criterion is provided, to the NEF by the AF and is associated with the capability of aggregating.9.A method performed by an application function (AF) performing vertical federated learning (VFL) in a wireless communication system, the method comprising:transmitting, to a network repository function (NRF) via a network exposure function (NEF), a request message for discovering at least one network data analytics function (NWDAF), wherein at least one network function (NF) profile of at least one NWDAF is registered to the NRF;receiving, from the NRF via the NEF, a response message corresponding to the request message including an external NWDAF ID of the discovered NWDAF, based on the request message;storing the external NWDAF ID included in the response message, wherein the discovered NWDAF is selected by the NEF for the VFL;transmitting, to the discovered NWDAF via the NEF, a VFL preparation request message; andreceiving, from the NWDAF via the NEF, a response message corresponding to the VFL preparation request message, wherein the response message is an aggregate of at least one response message for preparing the VFL transmitted from at least one other NWDAF to the discovered NWDAF.10.The method of claim 9, transmitting a request message for discovering at least one NWDAF further comprising;transmitting, to the NEF, a message for registering capability as a VFL server for an Analytics ID; andtransmitting the request message for discovering at least one NWDAF.11.The method of claim 9, receiving, from the NWDAF via the NEF, a response message corresponding to the VFL preparation request message;wherein a VFL preparation request message is transmitted, from the discovered NWDAF to the at least one other NWDAF and a response corresponding to the request is transmitted, from the at least one other NWDAF to the discovered NWDAF and is aggregated by the discovered NWDAF.12.The method of claim 9, further comprising;transmitting, to the discovered NWDAF via the NEF, a VFL preparation request message, wherein the external NWDAF ID of the discovered NWDAF is mapped to an internal NWDAF ID of the discovered NWDAF by the NEF and a generic public subscription identifier (GPSI) of a terminal is mapped to a subscription permanent identifier (SUPI) of the terminal by the NEF; andreceiving, from the discovered NWDAF via the NEF, a response message corresponding to the VFL preparation request message, wherein the internal NWDAF ID of the discovered NWDAF is mapped to the external NWDAF ID of the discovered NWDAF, by the NEF.13.An application function (AF) entity in a wireless communication system, the AF entity comprising:at least one transceiver;at least one processor communicatively coupled to the at least one transceiver; andat least one memory, communicatively coupled to the at least one processor, storing instructions executable by the at least one processor individually or in any combination to cause the AF entity to:transmit, to a network repository function (NRF) via a network exposure function (NEF), a request message for discovering at least one network data analytics function (NWDAF), wherein the at least one network function (NF) profile of at least one NWDAF is registered to the NRF;receive, from the NRF via the NEF, a response message corresponding to the request message including an external NWDAF ID of the discovered NWDAF, based on the request message;store the external NWDAF ID included in the response message, wherein the discovered NWDAF is selected by the NEF for the VFL;transmit, to the discovered NWDAF via the NEF, a VFL preparation request message; andreceive, from the NWDAF via the NEF, a response message corresponding to the VFL preparation request message, wherein the response message is an aggregate of at least one response message for preparing the VFL transmitted from at least one other NWDAF to the discovered NWDAF.14.The AF entity of claim 13, wherein the AF entity is further caused to;transmit, to the NEF, a message for registering capability as a VFL server for an Analytics ID; andtransmit the request message for discovering at least one NWDAF.15.The AF entity of claim 13, wherein the AF entity is caused totransmit, to the discovered NWDAF via the NEF, the VFL preparation request message, wherein the external NWDAF ID of the discovered NWDAF is mapped to an internal NWDAF ID of the discovered NWDAF by the NEF and a generic public subscription identifier (GPSI) of a terminal is mapped to a subscription permanent identifier (SUPI) of the terminal by the NEF; andreceive, from the NWDAF via the NEF, a response message corresponding to the VFL preparation request message, wherein the internal NWDAF ID of the discovered NWDAF is mapped to the external NWDAF ID of the discovered NWDAF, by the NEF.

Images