Security capability registration and negotiation between longitudinal federated learning participants

Abstract

The present disclosure relates, among other things, to a method comprising: - obtaining vertical federated learning, VFL, participant information representing at least one VFL participant associated with respective security capability information, wherein the security capability information indicates that the respective VFL participant supports at least one security mechanism; - determining that a preferred security mechanism has been specified; and - if the preferred security mechanism has been specified: - obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first selected VFL participant supports the preferred security mechanism; and - preparing a VFL process using the first selected VFL participant and the preferred security mechanism for training a machine learning model.

Description

Technical Field

[0001] This disclosure relates to, but is not limited to, longitudinal federated learning (VFL) in the context of communication systems defined by 3GPP standards (e.g., 5G standards, also known as New Radio (NR) standards). Specifically, this disclosure relates to security aspects related to the training of artificial intelligence (AI) / machine learning (ML) models using VFL. Background Technology

[0002] Artificial intelligence (AI) is a current technological field focused on creating systems that simulate human intelligent decision-making and / or behavior in specific domains. Machine learning (ML) is a subset of AI, focusing on statistical learning from training data. Artificial neural network (ANN) algorithms are examples of a class of algorithms used in ML, inspired by simplified neurons in the brain.

[0003] Vertical federated learning (VFL) is a federated learning setup in which multiple entities (i.e., the so-called VFL participants) can jointly train an ML model without exposing their original training data, local ML models, or all their local ML model parameters. For communication systems defined by the 3GPP standard, VFL has been introduced in 3GPP Release 19. Various use cases include VFLs between network functions (NFs) of the mobile core network, such as VFLs between Network Data Analysis Functions (NWDAFs), VFLs between Application Functions (AFs) and NWDAFs, or NWDAF-assisted VFLs between AFs. In this context, security aspects related to VFLs can be significant, especially in the case of cross-domain VFLs, such as VFLs involving the 5G core (5GC) network and AFs.

[0004] Furthermore, various entities may assume different VFL roles, which can include VFL initiators, VFL coordinators, active VFL participants, and passive VFL participants, and can be represented by various functional groups (NFs). In this respect, the VFL initiator can determine that the VFL process should be executed, and can participate in the discovery (i.e., search) of VFL participants and the preparation for VFL model training. The VFL coordinator may be required to participate in the VFL process. When participating, the VFL coordinator can be responsible for maintaining the VFL process, as well as authorizing or removing VFL participants. Maintaining the VFL process involves collecting the required inputs from a set of VFL participants in each iteration and providing them with the information needed to run the next iteration. Active VFL participants can be NFs that perform VFL model training using the required data labels. Passive VFL participants can be NFs that perform VFL model training without data labels; these are just a few non-limiting examples. Summary of the Invention

[0005] In a Virtual Framework (VFL), there are active participants who include or contain labels or benchmark truths of certain data, and / or passive participants who contain other important features that may be needed for ML model training. These features may not be available at the active participants. This can lead to security challenges, as the security of data and model parameters shared among VFL members can become extremely important. In this context, the term "benchmark truth," based on its conventional meaning in the field of machine learning and its use here, refers to the correct or "true" answer to a specific question or question. For example, in an image classification system, an algorithm can learn to classify each image into a set of categories, and the algorithm can be trained using training data with benchmark truth labels that indicate the true category of each image (e.g., manually verified by a human operator).

[0006] Currently, there is no mechanism in the 3GPP standard to prevent VFL participants from decrypting training data from model parameters and / or embeddings shared by other VFL participants. Here, the term "training data" refers to local data used by a VFL participant for training, consisting of labels (in the case of active participants) and other features (in the case of passive participants) and their corresponding values. Furthermore, no mechanism has been defined to detect what types of security capabilities a particular VFL participant supports and how the VFL coordinator uses this information when preparing or controlling the VFL process.

[0007] In view of the foregoing, exemplary embodiments of this disclosure may, among other things, allow for enhanced security in the context of VFL procedures being performed by an NF in a communication system. More specifically, exemplary embodiments of this disclosure may, among other things, allow for the coordination of security mechanisms among VFL participants.

[0008] According to an exemplary aspect of this disclosure, a method is disclosed. This method may include at least: - Obtain longitudinal federated learning (VFL) participant information associated with corresponding security capability information, representing at least one VFL participant, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism; - Determine the preferred security mechanism that has been specified; - If a preferred security mechanism has already been specified: - Obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism; and - Prepare the VFL process using the first selected VFL participants and the preferred security mechanism for training a machine learning model.

[0009] In an example embodiment of the method, the method may be performed by at least one network function (NF). For example, the at least one network function may be included by an apparatus. For example, the at least one network function may be at least one network function of a communication system (e.g., a communication system according to 3GPP standards). Furthermore, the at least one network function may be at least one network function of a 5GC network. For example, the at least one network function may be or include a Network Data Analysis Function (NWDAF). For example, the NWDAF may be or include one or more VFL participants. For example, the NWDAF may include a Model Training Logic Function (MTLF) that performs the method or is involved in the method. For example, the MTLF may be or include a VFL Initiator NF. For example, the MTLF may be or include a VFL Coordinator NF. For example, the at least one network function may be or include a Network Repository Function (NRF). For example, the at least one network function may be or include a Network Open Function (NEF). For example, the at least one network function may be or include an Application Function (AF). For example, the AF may be or include one or more (e.g., single or multiple, at least two) VFL participants. For example, the at least one network function may include a VFL Initiator NF, a VFL Coordinator NF, an NRF, a NEF, and one or more VFL participants. For example, the at least one network function may include a VFL initiator NF represented or included by the NWDAF's MTLF, a VFL coordinator NF represented or included by the NWDAF's MTLF, an NRF, a NEF, one or more VFL participants represented or included by the NWDAF, and one or more VFL participants represented or included by the AF. In an embodiment, the VFL coordinator NF and the NEF may be the same entity.

[0010] As used in this paper, VFL participants can be understood as entities capable of or intended to participate in the VFL process used to train machine learning (ML) models. In this context, artificial intelligence (AI) is a technological field focused on creating intelligent systems. Machine learning (ML) is a subset of AI, focusing on statistical learning from data. Artificial neural network (ANN) algorithms are examples of a class of algorithms used in ML, inspired by simplified neurons in the brain. ML is the ML lifecycle, which can be broken down into two parts: the training phase and the ML inference phase. In the training phase, the ML model is created or “trained” by running a specified subset of data into the model, and in the ML inference phase, the ML model is applied to the data to produce actionable outputs. Thus, a VFL participant can be understood as an entity capable of training its local ML model using a set of training data, sharing only some model parameters / embeddings (e.g., locally predicted labels) with other VFL participants after each training round, without sharing its original training data or local ML model with other VFL participants, such as using a VLF training framework like MMVFL (S. Feng et al., “MMVFL: A Simple Vertical Federated Learning Framework for Multi-Class Multi-Participant Scenarios”, published in Sensors, 24(2), 619).<URL: http: / / www.mdpi.com / 1424-8220 / 24 / 2 / 619> ).

[0011] As used herein, VFL participant information can be understood as any information or data that could identify at least one group of VFL participants and determine whether each VFL participant in that group supports at least one security mechanism. For example, VFL participant information may constitute or include a group or list of VFL participants associated with corresponding security capability information indicating whether at least one VFL participant supports at least one security mechanism. For example, VFL participant information may constitute or include a table in which a first column or row lists at least one VFL participant, and a second column or row contains information such as a true / false flag, a yes / no flag, binary information, etc., indicating whether the corresponding VFL participant in the first column or row supports at least one specific security mechanism. For example, the table may include one or more additional columns or rows containing information such as a true / false flag, a yes / no flag, binary information, etc., to indicate whether the corresponding VFL participant in the first column or row supports one or more additional specific security mechanisms.

[0012] As used in this paper, a security mechanism can be understood as any mechanism or method capable of preventing VFL participants from obtaining ML model training data from other VFL participants in plaintext (i.e., decrypted). For example, a security mechanism can be based on encryption, such as cryptographic defense strategies (CDS) using multi-party computation (MPC), trusted execution environments (TEEs), and / or homomorphic encryption (HE). For example, a security mechanism can be based on privacy protection, such as privacy protection through randomly added noise, gradient discretization, gradient sparsity, or a combination thereof. (See, for example, "Vertical Federated Learning: Concepts, Advances, and Challenges" by Yo Liu et al., 2024, published in the IEEE Transactions on Knowledge and Data Engineering).<URI: https: / / ieeexpore.ieee.org / document / 10415268> Examples of suitable security mechanisms based on CDS and / or privacy protection are provided in the document.

[0013] As used herein, security capability information can be understood as any information or data that, at least in part, may determine whether a VFL participant associated with a given security capability supports at least one specific security mechanism. For example, security capability information may constitute or include corresponding true / false or no flags for the respective VFL participant and for the respective specific security mechanism. For example, VFL participants and security capability information may be associated in the same row or column of a table or array. For example, VFL participants and security capability information may be associated using a bijective mathematical function that assigns the corresponding security capability information to at least one VFL participant.

[0014] As used herein, a preferred security mechanism can be understood as a security mechanism that is enforced for the VFL process. For example, the preferred security mechanism can be specified by the NF participating in the VFL process, such as the VFL initiator NF or the VFL coordinator NF. For example, one or more exceptions can be defined, wherein if the corresponding exception is applied, the use of the preferred security mechanism is not enforced.

[0015] As used herein, a first selection of a VFL participant among at least one VFL participant can be understood as a first subgroup obtained by selecting VFL participants from a group of VFL participants represented by VFL participant information. The first selection of a VFL participant may include zero, one, more than one, or all VFL participants from that group of VFL participants.

[0016] As used herein, a VFL participant that supports security mechanisms can be understood as one that enables VFL participants to execute VFL procedures using the appropriate security mechanisms.

[0017] As used herein, the VFL process can be understood as the training of a machine learning (ML) model using longitudinal federated learning (VFL). VFL can be understood as a federated learning setting where multiple participants with different features about the same set of users, or different features about overlapping samples, train an ML model without exposing their original data or model parameters. After training, the ML model can be used for inference. For example, the VFL process can be performed using an ML model training algorithm such as gradient descent. However, this disclosure is not limited to using gradient descent, and other ML model training algorithms are conceivable, such as distributed coordinate descent.

[0018] According to an exemplary aspect of this disclosure, the method may include obtaining VFL participant information representing at least one VFL participant associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism. In embodiments, the VFL participant information may represent more than one VFL participant, such as at least two VFL participants or a number of VFL participants. In embodiments, although the VFL participant information represents more than one (e.g., at least two) VFL participant, the corresponding security capability information may exist for (e.g., only) one VFL participant. In an exemplary embodiment, the corresponding security capability information may exist for more than one VFL participant, such as for each (or all) VFL participant represented by the VFL participant information. In embodiments, the security capability information may indicate whether the corresponding VFL participant supports at least one security mechanism. In an exemplary embodiment, the security capability information may indicate whether the corresponding VFL participant supports at least one security mechanism. In an example embodiment, security capability information may indicate (or, for example, whether or / no) that a corresponding VFL participant supports a first specific security mechanism, and (or, for example, whether or / no) that a corresponding VFL participant supports a second specific security mechanism. In an example embodiment, security capability information may indicate (or, for example, whether or / no) that a corresponding VFL participant supports at least one other specific security mechanism. In embodiments, obtaining VFL participant information may be performed by a VFL initiator NF, a VFL coordinator NF, an NRF, or any combination thereof. By obtaining VFL participant information representing at least one VFL participant associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism, information about the supported security mechanisms of the at least one VFL participant is provided to an NF of the communication system, based on this information, security mechanisms can be coordinated among VFL participants who will participate in the VFL process.

[0019] According to an exemplary aspect of this disclosure, the method may include determining that a preferred security mechanism has been designated. In embodiments, determining that a preferred security mechanism has been designated may include: determining whether a preferred security mechanism has been designated. In example embodiments, determining that a preferred security mechanism has been designated may include: determining whether a preferred security mechanism has been designated. In embodiments, determining that a preferred security mechanism has been designated may be performed by the VFL coordinator NF, NRF, or both. By determining that a preferred security mechanism has been designated, it is possible to coordinate security mechanisms among VFL participants who will be involved in the VFL process.

[0020] According to an exemplary aspect of the invention, the method may include obtaining or performing a first selection of VFL participants among at least one VFL participant if a preferred security mechanism has been specified, wherein the first-selected VFL participant supports the preferred security mechanism. In an example embodiment, obtaining or performing the first selection of VFL participants may be performed in response to determining that a preferred security mechanism has been specified. In an example embodiment, each of the first-selected VFL participants supports the preferred security mechanism. In an exemplary embodiment, obtaining or performing the first selection of VFL participants among at least one VFL participant may include: including all VFL participants in the first selection from the group of VFL participants represented by VFL participant information, for example, if it has been pre-determined that all of these VFL participants support the preferred security mechanism. In an embodiment, obtaining or performing the first selection of VFL participants may be performed by a VFL coordinator NF, wherein the VFL coordinator NF selects a first subgroup from the group of VFL participants represented by the VFL participant information. In an example embodiment, obtaining or executing a first selection of a VFL participant can be performed jointly by the VFL coordinator (NF) and the NRF (Non-RF), wherein the NRF selects a first subgroup from the groups of VFL participants represented by the VFL participant information, and wherein the NRF sends the first selection to the VFL coordinator (NF), which then obtains the first selection. By obtaining or executing a first selection of a VFL participant among at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism, it is possible to coordinate the preferred security mechanism among VFL participants who will participate in the VFL process.

[0021] According to an exemplary aspect of this disclosure, the method may include preparing a VFL process for training a machine learning model using a first-selected VFL participant and the preferred security mechanism, if a preferred security mechanism has been specified. In embodiments, the VFL preparation process may be performed in response to determining that a preferred security mechanism has been specified. In an exemplary embodiment, the VFL preparation process may be performed for jointly training a machine learning model, for example, if two or more VFL participants support the preferred security mechanism and therefore participate in the VFL process. In embodiments, the VFL preparation process may be performed by a VFL coordinator (NF) together with one or more VFL participants (e.g., the first-selected VFL participant). In an exemplary embodiment, the VFL initiator (NF) may participate in preparing the VFL process. In an exemplary embodiment, an NRF, NEF, or both may also participate. By using the first-selected VFL participant and the preferred security mechanism to prepare the VFL process for training a machine learning model, improvements in the security of the VFL process to be performed can be achieved.

[0022] According to an exemplary aspect of this disclosure, the method may include, if a preferred security mechanism has been specified, performing a VFL procedure for training a machine learning model using a first-selected VFL participant and the preferred security mechanism. This enables the execution of the VFL procedure and the training of the ML model via the VFL procedure.

[0023] In an example embodiment of the method, the method may further include: - If a preferred security mechanism has not yet been specified: - Identify a common security mechanism supported by the maximum number of VFL participants among the at least one VFL participant; - Obtain or execute a second selection of a VFL participant among the at least one VFL participant, wherein the second-selected VFL participant supports the common security mechanism; and - Prepare the VFL process using the second selected VFL participant and the common security mechanism for training a machine learning model.

[0024] As used herein, a common security mechanism can be understood as a security mechanism supported by the maximum number of VFL participants in a group of VFL participants represented by VFL participant information. In example embodiments, a common security mechanism may be supported by one, some, or all (e.g., each) of the VFL participants in the group.

[0025] In an example embodiment, identifying a common security mechanism, obtaining or executing a second selection of VFL participants, and preparing a VFL process for training a machine learning model using the second-selected VFL participants and the common security mechanism can be performed in response to determining that a preferred security mechanism has not yet been specified (e.g., after or subsequently after determination). In an example embodiment, identifying the common security mechanism can be performed by the VFL coordinator NF. In an example embodiment, each of the second-selected VFL participants supports the common security mechanism. In an exemplary embodiment, obtaining or executing a second selection of VFL participants among at least one VFL participant can include including all VFL participants in the group of VFL participants represented by VFL participant information in the second selection, for example, if it has been determined in advance that all of these VFL participants support the common security mechanism. In an example embodiment, obtaining or executing a second selection of VFL participants can be performed by the VFL coordinator NF, wherein the VFL coordinator NF selects a second subgroup from the group of VFL participants represented by VFL participant information. In an exemplary embodiment, for example, if two or more VFL participants support the common security mechanism and thus participate in the VFL process, a VFL preparation process can be performed for (e.g., jointly) training a machine learning model. In an embodiment, the VFL preparation process may be performed by a VFL coordinator (NF) together with one or more VFL participants (e.g., a second-selected VFL participant). In an exemplary embodiment, the VFL initiator (NF) may participate in preparing the VFL process. In an embodiment, an NRF, NEF, or both may also be included. By identifying the common security mechanism, obtaining or performing a second selection of VFL participants, and using the second-selected VFL participant and the common security mechanism to prepare the VFL process for training a machine learning model, it is possible to coordinate the common security mechanism among VFL participants who will participate in the VFL process. Therefore, the security of the VFL process to be performed can be improved.

[0026] According to an exemplary aspect of this disclosure, the method may include, if a common security mechanism has been specified, using the second selected VFL participant and the common security mechanism to perform a VFL process for training a machine learning model. This enables the execution of a VFL process and the training of an ML model via the VFL process.

[0027] In an example embodiment of the method, if (and for example, in response to a determination) a preferred security mechanism has been specified, and if (and for example, in response to a determination) additional conditions are met or not met, the process may also involve identifying a common security mechanism, obtaining or executing a second selection of VFL participants, and using the second-selected VFL participants and the common security mechanism to prepare (and / or execute) a VFL process for training a machine learning model. For example, if a preferred security mechanism has been specified, but one or more (e.g., greater than a threshold or a percentage greater than a threshold) VFL participants do not support the preferred security mechanism, the process may involve identifying a common security mechanism, obtaining or executing a second selection of VFL participants, and preparing (and / or executing) the corresponding VFL process.

[0028] In an example embodiment of the method, the method may include at least: - Obtain longitudinal federated learning (VFL) participant information associated with corresponding security capability information, representing at least one VFL participant, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism; - Determine the preferred security mechanism that has been specified; - If a preferred security mechanism has not yet been specified: - Identify a common security mechanism supported by the maximum number of VFL participants among the at least one VFL participant; - Obtain or execute a second selection of a VFL participant among the at least one VFL participant, wherein the second-selected VFL participant supports the common security mechanism; and - Prepare the VFL process using the second selected VFL participant and the common security mechanism for training a machine learning model.

[0029] In an example embodiment of the method, the method may further include: - Register the at least one VFL participant, wherein registering the corresponding VFL participant includes providing VFL participant identification information associated with the corresponding security capability information, wherein the VFL participant identification information indicates the corresponding VFL participant.

[0030] As used herein, registering a VFL participant can be understood as a process in which VFL participant identification information associated with corresponding security capability information is provided from a first entity (e.g., an NF) to a second entity (e.g., a different NF). For example, registering a VFL participant may include providing information identifying the VFL participant. For example, registering a VFL participant may include providing information indicating the VFL capabilities of the VFL participant. For example, registering a VFL participant may include providing information indicating the role of the VFL participant, such as the role of an active VFL participant or the role of a passive VFL participant. For example, registering a VFL participant may include providing information indicating at least one security mechanism supported by the VFL participant. For example, registering a VFL participant may include providing information indicating at least one parameter associated with at least one security mechanism, such as a parameter indicating the encryption strength (e.g., indicating the encryption key length) in an encryption-based security mechanism, or a desired minimum privacy threshold in a privacy-preserving security mechanism; these are merely non-limiting examples. In example embodiments, VFL participants may register in an NRF. For example, VFL participants implemented or included by an NWDAF or a trusted AF may register in an NRF. In an example embodiment, a VFL participant may register in the NEF, and the NEF may later register that VFL participant with the NRF. For example, a VFL participant implemented or included by an untrusted AF may register in the NEF, and the NEF may later register the VFL participant with the NRF. As used herein, VFL participant identification information can be understood as any information or data that can identify a particular VFL participant. By registering the at least one VFL participant—where registering the corresponding VFL participant includes providing VFL participant identification information associated with the corresponding security capability information, wherein the VFL participant identification information indicates the corresponding VFL participant—VFL participant information for coordinating security mechanisms among VFL participants can be obtained.

[0031] In an example embodiment of the method, the method may further include: - Request to initiate the VFL process.

[0032] In the example embodiment, the request to initiate a VFL process can be executed by the VFL initiating NF. In the example embodiment, the request to initiate a VFL process can be executed by the VFL initiating NF sending a corresponding request to the VFL coordinating NF. By requesting to initiate a VFL process, the VFL initiating NF can be enabled to initiate the VFL process. In the example embodiment, the request to initiate a VFL process may include an indication that the VFL process requires a security mechanism. For example, this indication could be in the form of a "securedVFL" flag. Therefore, the VFL initiating NF can ensure that a security mechanism is used to execute the VFL process.

[0033] In an exemplary embodiment of the method, obtaining the VFL participant information may include: - Request and / or perform discovery of VFL participants.

[0034] As used herein, discovery can be understood as a search for VFL participants that meet at least one requirement. For example, the at least one requirement may be or include that the relevant VFL participant is already registered with the NRF. For example, the at least one requirement may be or include that the relevant VFL participant is already registered with the NRF and supports at least one security mechanism. For example, the at least one requirement may be or include security-related requirements, such as the requirement that the VFL participant to be discovered supports a specified security mechanism (e.g., a preferred security mechanism), or the requirement that some parameters of the security mechanism meet certain conditions (e.g., equal to or greater than a threshold). As a non-limiting example, the condition may be conforming to the encryption strength (e.g., encryption key length) specified in an encryption-based security mechanism. As another non-limiting example, the threshold may be a minimum privacy threshold expected in a privacy-preserving security mechanism. In an exemplary embodiment, requesting and / or performing discovery of VFL participants may be performed by the VFL coordinator NF together with the NRF, wherein the VFL coordinator NF sends a discovery request to the NRF, and the NRF subsequently performs the corresponding discovery among the VFL participants registered with the NRF. In an exemplary embodiment, requesting and / or performing discovery of VFL participants can be performed by the VFL initiator NF in conjunction with the NRF, as the VFL initiator NF sends a discovery request to the NRF, and the NRF subsequently performs the corresponding discovery among the VFL participants registered with the NRF. By requesting and / or performing discovery of VFL participants, it is possible to identify VFL participants so that security mechanisms can be coordinated among them. In an exemplary embodiment, requesting discovery of VFL participants may include an indication that the VFL process requires a security mechanism. For example, this indication may be in the form of a “securedVFL” flag. For example, this indication may be or correspond to an indication included in a previous request to initiate the VFL process.

[0035] In an exemplary embodiment of the method, obtaining VFL participant information may include: - As a result of the discovery of VFL participants, obtain a list representing the at least one VFL participant.

[0036] In an example embodiment, as a result of a discovery performed by the NRF, the list is acquired by the NRF and then sent to the VFL coordinator NF that requested the discovery (and is therefore also acquired by the VFL coordinator NF). In another example embodiment, as a result of a discovery performed by the NRF, the list is acquired by the NRF and then sent to the VFL initiator NF that requested the discovery (and is therefore also acquired by the VFL initiator NF). In yet another example embodiment, the list received by the VFL initiator NF can be sent to the VFL coordinator NF (and is therefore also acquired by the VFL coordinator NF), for example, along with a request to initiate a VFL process. By acquiring a list representing at least one VFL participant, a security mechanism can be implemented that can coordinate among the VFL participants in the list who will participate in the VFL process.

[0037] In an exemplary embodiment of the method, obtaining VFL participant information may include: - For at least one VFL participant in the list, obtain security capability information associated with the corresponding VFL participant, wherein the corresponding security capability information is obtained together with the list.

[0038] Thus, the corresponding security capability information for at least one VFL participant in the list can be made available. In an example embodiment, the list including at least one VFL participant can be a column or row of a table or array, and the VFL participants can be associated with security capability information, specifically, they are all included in the same row (if the list including at least one VFL participant is a column) or the same column (if the list including at least one VFL participant is a row) of the table or array.

[0039] In an exemplary embodiment of the method, obtaining VFL participant information may include: - For at least one VFL participant whose associated security capability information has not yet been obtained along with the list, request the corresponding security capability information from the relevant VFL participant; and / or - Obtain the corresponding security capability information from the corresponding VFL participants.

[0040] Therefore, it is possible to make the security capability information available to at least one VFL participant whose associated security capability information has not yet been obtained along with the list. For example, at least one VFL participant whose associated security capability information has not yet been obtained along with the list could be a registered VFL participant without security capability information. Alternatively, at least one VFL participant whose associated security capability information has not yet been obtained along with the list could be a VFL participant whose security capability information has been registered but cannot be retrieved. In an example embodiment, requesting security capability information from a VFL participant can be performed by the VFL coordinator NF sending a corresponding request to the VFL participant. In an exemplary embodiment, obtaining security capability information from a corresponding VFL participant can be performed by the VFL coordinator NF receiving security capability information from the VFL participant, for example, as a response to a request for security capability information.

[0041] In an example embodiment of the method, the method may further include: - Initiate a security negotiation for the at least one VFL participant represented by the VFL participant information.

[0042] As used in this article, security negotiation can be understood as the process of identifying and coordinating security mechanisms among VFL participants.

[0043] In an example embodiment of the method, preparing the VFL process for training the machine learning model using the first selected VFL participant and the preferred security mechanism may include at least one of the following: - Finalize the selection of VFL participants; - Instantiate the VFL session identifier; - Configure the use of preferred security mechanisms; - Reject the VFL request.

[0044] Finalizing the VFL participant selection may include adding or removing individual VFL participants, or verifying the correctness of the VFL participant selection. In an example embodiment, configuring the use of a preferred security mechanism may include specifying security parameters associated with that preferred security mechanism. Rejecting a VFL request may be performed by the VFL coordinator (NF), for example, if (and in response to a determination) security negotiation has failed.

[0045] In an example embodiment of the method, using the second-selected VFL participant and the common security mechanism to prepare the VFL process for training a machine learning model may include at least one of the following: - Finalize the selection of VFL participants; - Instantiate the VFL session identifier; - Configure the use of common security mechanisms; - Reject the VFL request.

[0046] Finalizing the VFL participant selection may include adding or removing individual VFL participants, or verifying the correctness of the VFL participant selection. In embodiments, configuring the use of a common security mechanism may include specifying security parameters associated with that common security mechanism. Rejecting a VFL request may be performed by the VFL coordinator (NF), for example, if (and in response to a determination) security negotiation has failed.

[0047] In an exemplary embodiment of the method, requesting to initiate a VFL procedure may include specifying a preferred security mechanism. This allows the VFL procedure to be performed using the preferred security mechanism. In an exemplary embodiment, specifying a preferred security mechanism may include specifying at least one parameter associated with that preferred security mechanism. For example, the at least one parameter may define the encryption strength (e.g., encryption key length) in an encryption-based security mechanism. For example, the at least one parameter may define a desired minimum privacy threshold in a privacy-preserving security mechanism. In an exemplary embodiment, requesting the discovery of VFL participants may include specifying a pre-specified preferred security mechanism when requesting to initiate a VFL procedure. In an exemplary embodiment, performing the discovery of VFL participants may consider the pre-specified preferred security mechanism when requesting to discover VFL participants. For example, the discovery may consider the preferred security mechanism in a way that the discovery selects (e.g., selects only) VFL participants that support the preferred security mechanism. This allows the discovery to generate VFL participants that support the preferred security mechanism, thereby simplifying the execution or acquisition of a first selection of VFL participants. In an exemplary embodiment, performing the discovery of VFL participants may also consider at least one parameter associated with the preferred security mechanism. For example, the discovery may consider at least one parameter related to the preferred security mechanism, specifically by selecting (e.g., only selecting) VFL participants that meet the at least one parameter. Thus, the discovery can generate VFL participants that support the preferred security mechanism and meet the at least one parameter, thereby further simplifying the execution or acquisition of a first selection of VFL participants.

[0048] In an example embodiment of the method, at least one security mechanism is at least one of the following: - Security mechanisms based on encryption; - A security mechanism based on privacy protection.

[0049] For example, encryption-based security mechanisms can be based on homomorphic encryption or additive homomorphic encryption. For example, encryption-based security mechanisms can involve encrypting and / or decrypting data using an encryption key. For example, privacy-preserving security mechanisms can be based on different privacy considerations. For example, privacy-preserving security mechanisms can involve modifying at least one gradient used in stochastic gradient descent of an ML model. For example, modification of at least one gradient can involve adding noise. For example, privacy-preserving security mechanisms can be based on the Ipsilon parameter and / or the delta parameter.

[0050] In an example embodiment of the method, at least one parameter of at least one security mechanism may vary according to at least one of the following: - The identifier of the machine learning model to be trained using the VFL process; - VFL participants who participate in the VFL process.

[0051] For example, at least one security mechanism can vary depending on the granularity of the analytics identifier or the model identifier of the ML model to be trained. For example, the analytics ID used for advertising may have different privacy noise additions than the analytics ID used for network optimization. For example, this at least one security mechanism may differ from the case where (e.g., all) VFL participants are from the same vendor, compared to the case where the VFL combines participants from different vendors or different domains (e.g., 5G core and AF).

[0052] According to an exemplary aspect, a corresponding apparatus for network functions is disclosed. The apparatus may include at least one processor and at least one memory. The at least one memory may store instructions that, when executed by the at least one processor, cause the apparatus to perform at least the following: - Obtain longitudinal federated learning (VFL) participant information associated with corresponding security capability information, representing at least one VFL participant, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism; - Determine the preferred security mechanism that has been specified; - If a preferred security mechanism has already been specified: - Obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism; and - Prepare the VFL process using the first selected VFL participants and the preferred security mechanism for training a machine learning model.

[0053] In an exemplary aspect, the disclosure of any action of the method should also be considered as a disclosure of instructions that, when executed by at least one processor of the device, cause the device to perform the corresponding action.

[0054] In example embodiments, the network function may include or be included in a device. In example embodiments, the network function may be a VLF coordinator. In example embodiments, the network function may be a Network Data Analysis Function (NWDAF). In an exemplary embodiment, the NWDAF may include a Model Training Logic Function (MTLF).

[0055] According to an exemplary aspect, a corresponding apparatus is also disclosed. The apparatus according to the exemplary aspect may include components for performing the method according to the exemplary aspect. In the exemplary aspect, any disclosure of method actions should also be considered as a disclosure of components for performing the corresponding method actions. The apparatus according to the exemplary aspect may be an apparatus for network functionality.

[0056] In example embodiments, the network function may include or be included in a device. In example embodiments, the network function may be a VLF coordinator. In example embodiments, the network function may be a Network Data Analysis Function (NWDAF). In an exemplary embodiment, the NWDAF may include a Model Training Logic Function (MTLF).

[0057] The components or functions of the apparatus according to the exemplary aspects can be implemented in hardware and / or software. They may include one or more modules or units that provide the corresponding functions. For example, they may include at least one processor for executing computer program code to perform the desired function, at least one memory for storing the program code, or both. Alternatively, they may include circuitry, for example, designed to implement the desired function, such as implemented in a chipset or chip (e.g., an integrated circuit). Typically, the component may include, for example, one or more processing components or processors.

[0058] According to an exemplary aspect, a corresponding computer-readable medium is disclosed. The computer-readable medium may be non-volatile. The computer-readable medium according to the exemplary aspect may include program instructions that, when executed by a means for network functions, cause the means to perform the method according to the exemplary aspect.

[0059] For example, a computer-readable medium according to an exemplary aspect may be a disk, or a memory, or a similar medium. Program instructions may be stored in the computer-readable medium in the form of instructions encoded in the computer-readable medium. The computer-readable medium may be intended to participate in the operation of a device, such as internal or external memory, such as a computer's read-only memory (ROM) or hard disk, or may be intended for the distribution of programs, such as an optical disk.

[0060] According to an exemplary aspect, a corresponding computer program is disclosed. The computer program according to the exemplary aspect may include instructions that, when executed by a device, cause the device to perform the method according to the exemplary aspect. The device may be an apparatus for network functions.

[0061] A computer program according to the exemplary aspects may be stored on a computer-readable storage medium, particularly a tangible and / or non-volatile computer-readable storage medium. In particular, a computer program according to the exemplary aspects may be stored on a non-volatile computer-readable medium according to the exemplary aspects.

[0062] It should be understood that the embodiments disclosed herein are merely exemplary and not restrictive.

[0063] Other features of this disclosure will become apparent from the following detailed description taken in conjunction with the accompanying drawings. However, it should be understood that the drawings are intended for illustrative purposes only and not as definitions of limitations on this disclosure, which should be referenced to the appended claims. It should also be understood that the drawings are not drawn to scale and are intended only to conceptually illustrate the structures and processes described herein. Attached Figure Description

[0064] Some exemplary embodiments will now be described with reference to the accompanying drawings, in which, Figure 1 The schematic diagram exemplarily illustrates the system architecture of a communication system in which exemplary embodiments of the present disclosure may be performed; Figure 2 An example embodiment of the method according to this disclosure is illustrated with a data flow diagram; Figure 3 Another example embodiment of the method according to this disclosure is illustrated with a data flow diagram; Figure 4 Another example embodiment of the method according to this disclosure is illustrated in a flowchart; Figure 5 An example embodiment of a device for network functions according to this disclosure is illustrated in schematic block diagram; and Figure 6 An example of a tangible and non-volatile computer-readable storage medium is illustrated schematically. Detailed Implementation

[0065] The following description is intended to enhance understanding of this disclosure and should be understood to complement and be read in conjunction with the description of exemplary embodiments of this disclosure provided in the foregoing summary of the invention section of this specification.

[0066] In the following text, see references Figure 1This disclosure describes the system architecture of a communication system in which the present disclosure can be applied. While the specific systems in the following examples are directed to 5G systems, they are to be considered (e.g., only) as non-limiting examples, and any next-generation systems or standards above and beyond 5G are also considered.

[0067] The 5G system (5GS) 100 is a communication system (e.g., a 3GPP system) that includes a (radio) access network ((R)AN) 102 (generally referred to here as RAN) and a 5G core network (5GC) that communicate with 5G user equipment (UE) 101. RAN 102 and 5GC together can be referred to as a 5G network, a 5G mobile network, a 5G communication network, etc.

[0068] RAN 102 provides radio or wireless connectivity to UE 101 and connects UE 101 to 5GC. RAN 102 may include Next Generation Radio Access Network (NG-RAN), Evolved Universal Terrestrial Radio Access Network (E-UTRAN), Non-3GPP Access Network (N3AN), Non-Terrestrial Access Network (NTN), and / or other types of RAN connected to 5GC. RAN 102 can support access through at least one RAN node, such as a gNodeB (gNB), ng-eNodeB (ng-eNB), eNodeB (eNB), and / or a Wireless Local Area Network (WLAN) access point. RAN 102 may support satellite radio access, new radio access technologies (RAT), etc.

[0069] The 5G core network (5GC) includes network functions (NFs), which can be implemented as network elements on dedicated hardware, chips or chipsets included in network elements, software instances running on appropriate dedicated or general-purpose hardware, virtualized network functions (VNFs) instantiated on dedicated or general-purpose virtualization platforms, etc. Some or all of these NFs can be stateless, virtualized, hosted on virtual machines or servers, operating system-level virtualized, containerized, hosted in multi-user space, and can be application virtualized, network virtualized, storage virtualized, server virtualized, or desktop virtualized. They can be hosted / stored in / at a cloud computing environment, etc.

[0070] like Figure 1The NFs included in the 5GC shown are: Access and Mobility Management Function (AMF) 109, Session Management Function (SMF) 110, User Plane Function (UPF) 112 (connectable to an external data network (DN) 113), Policy Control Function (PCF) 105, Authentication Server Function (AUSF) 108, Unified Data Management (UDM) 106, Application Function (AF) 107, Network Open Function (NEF) 104, Network Repository Function (NRF) 103, and Network Data Analysis Function (NWDAF) 111, including Model Training Logic Function (MTLF) 111a and Analysis Logic Function (AnLF) 111b. Typically, the 5GC may include functions not mentioned here and... Figure 1 Other NFs not shown, such as Network Slice Selection Function (NSSF), or NFs not mentioned above.

[0071] Now go to Figure 2 This diagram illustrates a data flow diagram of an example embodiment of the method according to the present disclosure. More specifically, it illustrates a series of actions and / or steps between VFL initiator NF 201, VFL coordinator NF 202, NRF 203, NEF 204, first VFL participant 205, and second VFL participant 206. In this non-limiting example, VFL initiator NF 201 and VFL coordinator NF 202 are implemented or included by the Model Training Logic Function (MTLF) of NWDAF, but they can typically be implemented or included by other NFs. Furthermore, although two VFL participants are involved in this exemplary embodiment, the present disclosure generally covers any reasonable number of VFL participants, particularly one VFL participant or more than two VFL participants. Figure 2 In the example, the first VFL participant 205 is an active VFL participant and is implemented by NWDAF, while the second VFL participant 206 is also an active VFL participant but is implemented by AF. However, in general, any VFL participant involved can be active or passive and can be implemented by NWDAF, AF, or other NFs.

[0072] Within action 211, the first VFL participant 205 registers its VFL capabilities, its role as an active VFL participant, and the security mechanisms it supports in NRF 203. The supported security mechanisms can be, for example, cryptographically based security mechanisms, privacy-preserving security mechanisms, or both. Furthermore, the first VFL participant 205 indicates security-related parameters, such as a minimum privacy threshold desired in a privacy-preserving security mechanism (this is just one non-limiting example).

[0073] Within action 212, the second VFL participant 206 registers its VFL capabilities, its role as an active VFL participant, and the security mechanisms it supports in NEF 204. Similarly, the supported security mechanisms can be, for example, cryptographic security mechanisms, privacy-preserving security mechanisms, or both.

[0074] Within action 213, NEF 204 registers the VFL profile of the second VFL participant 206 implemented by AF in NRF 203. More specifically, NEF 204 registers in NRF 203 information previously (i.e., within action 212) registered in NEF 204 by the second VFL participant 206. This information specifically includes the security mechanisms supported by the second VFL participant 206.

[0075] Therefore, actions 211, 212, and 213 can be considered as registration of VFL participants. As a result of actions 211, 212, and 213, NRF 203 specifically possesses information identifying the first VFL participant 205 and the second VFL participant 206, as well as information indicating the security mechanisms supported by the first VFL participant 205 and the second VFL participant 206. Therefore, NRF 203 also possesses information on whether the first VFL participant 205 and the second VFL participant 206 each support at least one security mechanism.

[0076] Within action 220, the VFL initiator NF 201 sends a request to the VFL coordinator NF 202 to initiate the VFL procedure. This request includes the “securedVFL” flag, indicating that a security mechanism is required for the VFL procedure. Furthermore, the request specifies the preferred security mechanism.

[0077] Within action 230, the VFL coordinator NF 202, based at least in part on a request received from the VFL initiator NF 201 within action 220, sends a request to NRF 203 to discover suitable VFL participants for the VFL process. In this context, a suitable VFL participant can be understood as one that, among other things, sets the "securedVFL" flag, i.e., a VFL participant that supports at least one security mechanism.

[0078] Within action 240, for example, after performing the corresponding discovery of VFL participants supporting at least one security mechanism, NRF 203 sends a list of VFL participants to VFL coordinator NF 202. This list includes VFL participants supporting at least one security mechanism. In this non-limiting example, assume that first VFL participant 205 supports both cryptographic-based and privacy-preserving security mechanisms. Second VFL participant 206 supports (e.g., only) cryptographic-based security mechanisms. Since both first VFL participant 205 and second VFL participant 206 support at least one security mechanism, and since they are both registered with NRF 203, they are both found in the discovery process. Therefore, the list of VFL participants selected here includes first VFL participant 205 and second VFL participant 206. NRF 203 sends information associated with the first and second VFL participants along with the list of selected VFL participants, indicating the security mechanisms supported by the first and second VFL participants respectively. In this specific example, NRF 203 sends a form to VFL coordinator NF 202. This form includes a first column listing the VFL participants, a second column indicating whether each VFL participant supports cryptographic-based security mechanisms, and a third column indicating whether each VFL participant supports privacy-preserving security mechanisms. Given the security mechanisms supported by the first and second VFL participants specified above for this non-limiting example, the form may have the following exemplary structure and content:

[0079] As a result, the VFL coordinator NF 202 also possesses information to identify the first VFL participant 205 and the second VFL participant 206, as well as information indicating the security mechanisms supported by the first VFL participant 205 and the second VFL participant 206. Therefore, the VFL coordinator NF 202 also possesses information on whether the first VFL participant 205 and the second VFL participant 206 each support at least one security mechanism.

[0080] Within action 250, VFL coordinator NF 202 initiates a security negotiation with the first VFL participant 205 and the second VFL participant 206.

[0081] Actions 211, 212, 213, 220, 230, and 240 together can be considered as the registration and discovery phase, followed by the selection and security negotiation phase, which includes actions 260 and 270, as described below.

[0082] Within action 260, VFL coordinator NF 202 determines that VFL initiator NF 201 has specified a preferred security mechanism within action 220. Therefore, VFL coordinator NF 202 filters the list of VFL participants (or, in this specific example, a table) to identify VFL participants that support the preferred security mechanism. In doing so, VFL coordinator NF 202 performs a (first) selection of VFL participants that support the preferred security mechanism. In this non-limiting example, the preferred security mechanism specified by VFL initiator NF 201 within action 220 is a cryptographic security mechanism that is supported by both first VFL participant 205 and second VFL participant 206. Therefore, the (first) selection here includes both first VFL participant 205 and second VFL participant 206.

[0083] In an embodiment, the VFL coordinator NF 202 may specify a preferred security mechanism along with the discovery request of action 230. In this case, filtering the list (or table) of VFL participants to identify those supporting the preferred security mechanism may have already been performed by NRF 203. The resulting (e.g., first) selection is then sent to the VFL coordinator NF 202, which thereby obtains the (e.g., first) selection of VFL participants supporting the preferred security mechanism.

[0084] Within action 270, the VFL coordinator NF 202 or the VFL initiator NF 201 assigns a VFL session identifier, coordinates with the first VFL participant 205 and the second VFL participant 206 to prepare for the VFL process for training a machine learning model, and configures the use of the preferred security mechanism (i.e., an encryption-based security mechanism in this example).

[0085] In an embodiment, for example in the case of a cross-domain VFL including 5GC and AF, the VFL coordinator NF 202 and NEF204 can be the same entity.

[0086] exist Figure 2 Within the actions not shown, the VFL process for training the machine learning model can be executed by the first VFL participant 205 and the second VFL participant 206.

[0087] Now go to Figure 3 This illustrates a data flow diagram of another example embodiment of the method according to the present disclosure. Figure 2 similar, Figure 3 This illustrates a series of actions between VFL initiator NF 301, VFL coordinator NF 302, NRF 303, NEF 304, first VFL participant 305, and second VFL participant 306. Figure 3In this non-limiting example, the VFL initiator NF 301 and the VFL coordinator NF 302 are implemented or included by the Model Training Logic Function (MTLF) of the NWDAF, but they can typically be implemented or included by other NFs. Furthermore, although two VFL participants are involved in this exemplary embodiment, this disclosure generally covers any reasonable number of VFL participants, particularly one VFL participant or more than two VFL participants. Figure 3 In the example, the first VFL participant 305 is also an active VFL participant and is implemented by NWDAF, while the second VFL participant 306 is also an active VFL participant, but is implemented by AF. However, typically, any VFL participant involved can be active or passive and can be implemented by NWDAF, AF, or other NFs.

[0088] Within action 311, the first VFL participant 305 registers its VFL capabilities and its role as an active VFL participant in NRF 303, for example, but does not register the security mechanisms it supports.

[0089] Within action 312, the second VFL participant 306 registers its VFL capabilities and its role as an active VFL participant in NEF 304, for example, but does not register the security mechanisms it supports.

[0090] Within action 313, NEF 304 registers the VFL profile of the second VFL participant 306, implemented by AF, in NRF 303. More specifically, NEF 304 registers in NRF 303 information previously (i.e., within action 312) registered in NEF 304 by the second VFL participant 306. However, this information does not include the security mechanisms supported by the second VFL participant 306.

[0091] Therefore, actions 311, 312, and 313 can be considered as registration of VFL participants. As a result of actions 311, 312, and 313, NRF 303 possesses information identifying the first VFL participant 305 and the second VFL participant 306. However, with Figure 2 Unlike the exemplary implementation, the NRF 303 in this state does not have information indicating the security mechanisms supported by the first VFL participant 305 and the second VFL participant 306, or information indicating whether the first VFL participant 305 or the second VFL participant 306 supports at least one security mechanism.

[0092] Within action 320, the VFL initiator NF 301 sends a request to the VFL coordinator NF 302 to initiate the VFL procedure. This request includes the “securedVFL” flag, indicating that the VFL procedure requires a security mechanism. However, in Figure 3 In the example embodiments, with Figure 2 Unlike the example implementations, this request does not specify a preferred security mechanism.

[0093] Within action 330, VFL coordinator NF 302, based on a request received from VFL initiator NF 301 within action 320, sends a request to NRF 303 for the discovery of appropriate VFL participants in the VFL process. Specifically, in this context, appropriate VFL participants can be understood as VFL participants. These VFL participants may, among other things, set the "securedVFL" flag, meaning they support at least one security mechanism.

[0094] Within action 340, after performing the corresponding discovery, NRF 303 sends a list of VFL participants to VFL coordinator NF 302. In this exemplary embodiment, if NRF 303 in this state does not possess information indicating the security mechanisms supported by the registered VFL participants (see above), the discovery cannot be targeted at VFL participants supporting at least one security mechanism. This is also the case if a supported security mechanism has been registered but cannot yet be retrieved from NRF 303. Therefore, the discovery can be targeted at other criteria, such as VFL capabilities pre-registered in actions 311, 312, and 313. In this example, the discovery results in a first VFL participant 305 and a second VFL participant 306, derived from the analysis of their VFL capabilities. Therefore, the list of VFL participants sent from NRF 303 to VFL coordinator NF 302 includes first VFL participant 305 and second VFL participant 306, but this list does not contain information associated with the first and second VFL participants indicating the security mechanisms they support respectively. Therefore, in this example, VFL coordinator NF 302 retrieves this information separately from the first and second VFL participants in actions 345, 346, 347, and 348, as described below.

[0095] In addition, actions 311, 312, 313, 320, 330 and / or 340 can be considered as a registration and discovery phase, followed by a selection and security negotiation phase including actions 345, 346, 347, 348, 360 and / or 370, as described below.

[0096] Within action 345, VFL coordinator NF 302 requests information from first VFL participant 305 regarding the security mechanism supported by first VFL participant 305.

[0097] Within action 346, the first VFL participant 305 sends information to the VFL coordinator NF 302 indicating the relevant information of the security mechanism supported by the first VFL participant 305.

[0098] Within action 347, VFL coordinator NF 302 requests information from second VFL participant 306 regarding the security mechanisms supported by second VFL participant 306. In an embodiment, this request may be sent via NEF 304.

[0099] Within action 348, the second VFL participant 306 sends information to the VFL coordinator NF 302 indicating the appropriate security mechanism supported by the second VFL participant 306. In an exemplary embodiment, this information may be sent via NEF 304.

[0100] As a result, VFL coordinator NF 302 now possesses information identifying the first VFL participant 305 and the second VFL participant 306, as well as information indicating the security mechanisms supported by the first VFL participant 305 and the second VFL participant 306. Therefore, VFL coordinator NF 302 also possesses information on whether the first VFL participant 305 and the second VFL participant 306 each support at least one security mechanism. In this non-limiting example, it is again assumed that the first VFL participant 305 supports both encryption-based and privacy-preserving security mechanisms, while the second VFL participant 306 supports (e.g., only) encryption-based security mechanisms.

[0101] Within action 360, VFL coordinator NF 302 determines that a preferred security mechanism has not yet been specified, specifically since VFL initiator NF 301 has not specified one within action 320. Therefore, VFL coordinator NF 302 identifies the cryptographically based security mechanism as a common security mechanism, i.e., the type of security mechanism supported by the largest number of VFL participants. This is because, as stated above, in this non-limiting example, it is assumed that the cryptographically based security mechanism is supported by both the first VFL participant 305 and the second VFL participant 306, while it is assumed that the privacy-preserving security mechanism is supported (e.g., only) by the first VFL participant 305. Subsequently, VFL coordinator NF 302 filters the list of VFL participants to identify those supporting the common security mechanism. In doing so, VFL coordinator NF 302 performs a (second) selection of VFL participants supporting the common security mechanism (note that there is no corresponding first selection in this example embodiment because no preferred security mechanism is specified). In this non-restrictive example, the (second) choice includes a first VFL participant 305 and a second VFL participant 306, both of whom support the common security mechanism.

[0102] Within action 370, the VFL coordinator NF 302 or the VFL initiator NF 301 assigns a VFL session identifier, coordinates with the first VFL participant 305 and the second VFL participant 306 to prepare for the VFL process for training a machine learning model, and configures the use of the common security mechanism (i.e., an encryption-based security mechanism in this example).

[0103] In an embodiment, for example in the case of a cross-domain VFL including 5GC and AF, the VFL coordinator NF 302 and NEF304 can be the same entity.

[0104] exist Figure 3 Within the actions not shown, the VFL process for training the machine learning model can be executed by the first VFL participant 305 and the second VFL participant 306.

[0105] Now go to Figure 4 The flowchart shown is another example embodiment of the method according to this disclosure.

[0106] In action 410, VFL participant information representing at least one VFL participant is obtained in association with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism.

[0107] In action 420, the determination of the preferred security mechanism is performed.

[0108] In action 430, if a preferred security mechanism has been specified, a first selection of a VFL participant among at least one VFL participant is obtained or executed, wherein the first selected VFL participant supports the preferred security mechanism. Furthermore, the VFL process is prepared using the first selected VFL participant and the preferred security mechanism for training a machine learning model.

[0109] Figure 5 This is a block diagram of an example embodiment of a device 500 for network functions. For example, device 500 may be configured to perform one or more embodiments of the methods according to this disclosure (e.g., in...). Figure 2 , Figure 3 and Figure 4 The actions of the method (an embodiment of the method explained in the context of the example).

[0110] The device 500 includes a processor 501. The processor 501 may represent a single processor, or two or more processors, which are at least partially coupled, for example, via a bus. The processor 501 executes program code stored in program memory 502 (e.g., when executed on the processor 501, causing the device 500, individually and / or with one or more other devices 500, to execute program code according to one or more embodiments of the methods of this disclosure or portions thereof), and interfaces with main memory 503.

[0111] Program memory 502 may also include an operating system for processor 501. Some or all of memories 502 and 503 may also be included in processor 501. Program memory 502 and main memory 503 may be formed from a single memory.

[0112] Furthermore, the processor 501 can control the communication interface 504, which is configured, for example, to communicate according to a communication system (e.g., a 3GPP communication system). The communication interface 504 of the device 500 can be implemented by, for example, one or more network interfaces.

[0113] Components 502, 503 and 504 of device 500 may be connected to processor 501, for example, by means of one or more serial and / or parallel buses.

[0114] It should be understood that device 500 may include various other components.

[0115] Figure 6 This is a schematic diagram of an example of a tangible and non-volatile computer-readable storage medium according to the present disclosure, which can be used, for example, to implement Figure 5 The memory 502. Therefore, Figure 6 The image shows a flash memory 600 (which can be soldered or bonded to a printed circuit board, for example), a solid-state drive 601 (including multiple memory chips, such as flash memory chips), a magnetic hard disk drive 602, a secure digital (SD) card 603, a universal serial bus (USB) memory stick 604, an optical storage medium 605 (such as a CD-ROM or DVD), and a magnetic storage medium 606.

[0116] The following embodiments are also disclosed:

[0117] Example 1: A method comprising: - Obtain longitudinal federated learning VFL participant information representing at least one VFL participant associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism. - Determine the preferred security mechanism that has been specified; and - If a preferred security mechanism has already been specified: - Obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism; and - Prepare the VFL process using the first selected VFL participants and the preferred security mechanism for training a machine learning model.

[0118] Example 2: The method according to Example 1 further includes: - If a preferred security mechanism has not yet been specified: - Identify a common security mechanism supported by the maximum number of VFL participants among the at least one VFL participant; - Obtain or execute a second selection of a VFL participant among the at least one VFL participant, wherein the second-selected VFL participant supports the common security mechanism; and - Prepare the VFL process using the second selected VFL participant and the common security mechanism for training a machine learning model.

[0119] Example 3: The method according to Example 1 or 2 further includes: - Register the at least one VFL participant, wherein registering the corresponding VFL participant includes providing VFL participant identification information associated with the corresponding security capability information, wherein the VFL participant identification information indicates the corresponding VFL participant.

[0120] Example 4: The method according to any one of Examples 1 to 3 further includes: - Request to initiate the VFL process.

[0121] Example 5: According to any one of Examples 1 to 4, the method of obtaining the VFL participant information includes: - Request and / or perform discovery of VFL participants.

[0122] Example 6: According to the method described in Example 5, obtaining the VFL participant information includes: - As a result of the discovery of VFL participants, obtain a list representing the at least one VFL participant.

[0123] Example 7: According to the method described in Example 6, obtaining the VFL participant information includes: - For at least one VFL participant in the list, obtain security capability information associated with the corresponding VFL participant, wherein the corresponding security capability information is obtained together with the list.

[0124] Example 8: According to the method described in Example 6 or 7, obtaining the VFL participant information includes: - For at least one VFL participant in the list whose associated security capability information has not yet been obtained along with the list, request the corresponding security capability information from the relevant VFL participant; and / or - Obtain the corresponding security capability information from the corresponding VFL participants.

[0125] Example 9: The method according to any one of Examples 1 to 8 further includes: - Initiate a security negotiation for the at least one VFL participant represented by the VFL participant information.

[0126] Example 10: The method according to any one of Examples 1 to 9, wherein preparing the VFL process for training a machine learning model using the first selected VFL participant and the preferred security mechanism includes at least one of the following: - Finalize the selection of VFL participants; - Instantiate the VFL session identifier; - Configure the use of the preferred security mechanisms.

[0127] Example 11: According to the method of Example 2, wherein preparing the VFL process for training a machine learning model using the second selected VFL participant and the common security mechanism includes at least one of the following: - Finalize the selection of VFL participants; - Instantiate the VFL session identifier; - Configure the use of the aforementioned common security mechanism.

[0128] Example 12: According to the method described in Example 4, the request to initiate the VFL process includes: specifying the preferred security mechanism.

[0129] Example 13: The method according to any one of Examples 1 to 12, wherein the at least one security mechanism is at least one of the following: - Security mechanisms based on encryption; - A security mechanism based on privacy protection.

[0130] Example 14: The method according to any one of Examples 1 to 13, wherein at least one parameter of the at least one security mechanism varies according to at least one of the following: - Identifier of the machine learning model that will be trained using the VFL process; - VFL participants involved in the VFL process.

[0131] Example 15: The method according to any one of Examples 1 to 14, wherein the method is performed by at least one network function.

[0132] Example 16: An apparatus for network functions, the apparatus comprising at least one processor and at least one memory, the at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus to perform at least the following: - Obtain longitudinal federated learning VFL participant information representing at least one VFL participant associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism. - Determine the preferred security mechanism that has been specified; and - If a preferred security mechanism has already been specified: - Obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism; and - Prepare the VFL process using the first selected VFL participants and the preferred security mechanism for training a machine learning model.

[0133] Example 17: The apparatus according to Example 16, wherein the at least one memory further stores instructions that, when executed by at least one processor, cause the apparatus to perform: - If a preferred security mechanism has not yet been specified: - Identify a common security mechanism supported by the maximum number of VFL participants among the at least one VFL participant; - Obtain or execute a second selection of a VFL participant among the at least one VFL participant, wherein the second-selected VFL participant supports the common security mechanism; and - Prepare the VFL process using the second selected VFL participant and the common security mechanism for training a machine learning model.

[0134] Example 18: The apparatus according to Example 16 or 17, wherein the at least one memory further stores instructions that, when executed by the at least one processor, cause the apparatus to perform: - Register the at least one VFL participant, wherein registering the corresponding VFL participant includes providing VFL participant identification information associated with the corresponding security capability information, wherein the VFL participant identification information indicates the corresponding VFL participant.

[0135] Example 19: The apparatus according to any one of Examples 16 to 18, wherein the at least one memory further stores instructions that, when executed by the at least one processor, cause the apparatus to perform: - Request to initiate the VFL process.

[0136] Example 20: According to any one of Examples 16 to 19, obtaining the VFL participant information includes: - Request and / or perform discovery of VFL participants.

[0137] Example 21: According to the apparatus of Example 20, obtaining the VFL participant information includes: - As a result of the discovery of VFL participants, obtain a list representing the at least one VFL participant.

[0138] Example 22: According to the apparatus of Example 21, obtaining the VFL participant information includes: - For at least one VFL participant in the list, obtain security capability information associated with the corresponding VFL participant, wherein the corresponding security capability information is obtained together with the list.

[0139] Example 23: The apparatus according to Example 21 or 22, wherein obtaining the VFL participant information includes: - For at least one VFL participant in the list whose associated security capability information has not yet been obtained along with the list, request the corresponding security capability information from the relevant VFL participant; and / or - Obtain the corresponding security capability information from the corresponding VFL participants.

[0140] Example 24: The apparatus according to any one of Examples 16 to 23, wherein the at least one memory further stores instructions that, when executed by the at least one processor, cause the apparatus to perform: - Initiate a security negotiation for the at least one VFL participant represented by the VFL participant information.

[0141] Implementation 25: The apparatus according to any one of Embodiments 16 to 24, wherein preparing the VFL process for training a machine learning model using the first selected VFL participant and the preferred security mechanism includes at least one of the following: - Finalize the selection of VFL participants; - Instantiate the VFL session identifier; - Configure the use of the preferred security mechanisms.

[0142] Example 26: The apparatus according to Example 17, wherein preparing the VFL process for training a machine learning model using the second selected VFL participant and the common security mechanism includes at least one of the following: - Finalize the selection of VFL participants; - Instantiate the VFL session identifier; - Configure the use of the aforementioned common security mechanism.

[0143] Example 27: The apparatus according to Example 19, wherein requesting to initiate the VFL process includes: specifying the preferred security mechanism.

[0144] Example 28: The apparatus according to any one of Examples 16 to 27, wherein the at least one security mechanism is at least one of the following: - Security mechanisms based on encryption; - A security mechanism based on privacy protection.

[0145] Example 29: The apparatus according to any one of Examples 16 to 28, wherein at least one parameter of the at least one security mechanism varies according to at least one of the following: - Identifier of the machine learning model that will be trained using the VFL process; - VFL participants involved in the VFL process.

[0146] Example 30: The apparatus according to any one of Examples 16 to 29, wherein the apparatus is an apparatus for at least one network function.

[0147] Example 31: The apparatus according to any one of Examples 16 to 30, wherein the network function includes the apparatus or is included in the apparatus.

[0148] Example 32: The apparatus according to any one of Examples 16 to 31, wherein the network function is a VLF coordinator.

[0149] Example 33: The apparatus according to any one of Examples 16 to 32, wherein the network function is a network data analysis function (NWDAF) including a model training logic function (MTLF).

[0150] Example 34: An apparatus for network functions, the apparatus comprising: - A component for obtaining longitudinal federated learning VFL participant information representing at least one VFL participant associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism. - Components used to determine the preferred security mechanism; and - If a preferred security mechanism has already been specified: - Components for obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism; and - A component for preparing a VFL process for training a machine learning model using the first selected VFL participants and the preferred security mechanism.

[0151] Example 35: The apparatus according to Example 34 further includes: - If a preferred security mechanism has not yet been specified: - A component for identifying a common security mechanism supported by the maximum number of VFL participants among the at least one VFL participant; - Components for obtaining or performing a second selection of a VFL participant among the at least one VFL participant, wherein the second-selected VFL participant supports the common security mechanism; and - A component for preparing a VFL process for training a machine learning model using the second selected VFL participant and the common security mechanism.

[0152] Example 36: The apparatus according to Example 34 or 35 further includes: - A component for registering the at least one VFL participant, wherein registering a corresponding VFL participant includes providing VFL participant identification information associated with corresponding security capability information, wherein the VFL participant identification information indicates the corresponding VFL participant.

[0153] Implementation method 37: The apparatus according to any one of embodiments 34 to 36 further includes: - The component used to request the initiation of the VFL procedure.

[0154] Example 38: The apparatus according to any one of Examples 34 to 37, wherein the component for obtaining the VFL participant information includes: - A component used to request and / or perform discovery of VFL participants.

[0155] Example 39: The apparatus according to Example 38, wherein the component for obtaining the VFL participant information includes: - A component for obtaining a list representing the at least one VFL participant as a result of the discovery of VFL participants.

[0156] Example 40: The apparatus according to Example 39, wherein the component for obtaining the VFL participant information includes: - A component for obtaining security capability information associated with at least one VFL participant in the list, wherein the security capability information is obtained together with the list.

[0157] Example 41: The apparatus according to Example 39 or 40, wherein the component for obtaining the VFL participant information includes: - A component for requesting security capability information from at least one VFL participant in the list whose associated security capability information has not yet been obtained along with the list; and / or - A component for obtaining the corresponding security capability information from the corresponding VFL participants.

[0158] Implementation Scheme 42: The apparatus according to any one of Embodiments 34 to 41 further includes: - A component for initiating security negotiations for at least one VFL participant represented by the VFL participant information.

[0159] Implementation Scheme 43: The apparatus according to any one of Embodiments 34 to 42, wherein the components for preparing the VFL process for training a machine learning model using the first selected VFL participant and the preferred security mechanism include at least one of the following: - Components used to finalize the selection of VFL participants; - A component used to instantiate VFL session identifiers; - Components used to configure the preferred security mechanism.

[0160] Implementation 44: The apparatus according to Embodiment 35, wherein the components for preparing the VFL process for training a machine learning model using the second-selected VFL participant and the common security mechanism include at least one of the following: - Components used to finalize the selection of VFL participants; - A component used to instantiate VFL session identifiers; - Components used to configure the use of the aforementioned common security mechanism.

[0161] Example 45: The apparatus according to Example 37, wherein the component for requesting to initiate the VFL process includes: a component for specifying the preferred security mechanism.

[0162] Example 46: The apparatus according to any one of Examples 34 to 45, wherein the at least one security mechanism is at least one of the following: - Security mechanisms based on encryption; - A security mechanism based on privacy protection.

[0163] Example 47: The apparatus according to any one of Examples 34 to 46, wherein at least one parameter of the at least one security mechanism varies according to at least one of the following: - Identifier of the machine learning model that will be trained using the VFL process; - VFL participants involved in the VFL process.

[0164] Implementation Scheme 48: The apparatus according to any one of Embodiments 34 to 47, wherein the apparatus is an apparatus for at least one network function.

[0165] Implementation Scheme 49: The apparatus according to any one of Embodiments 34 to 48, wherein the network function includes the apparatus or is included in the apparatus.

[0166] Example 50: An apparatus according to one of Examples 34 to 49, wherein the network function is a VLF coordinator.

[0167] Example 51: The apparatus according to any one of Examples 34 to 50, wherein the network function is a network data analysis function (NWDAF) including a model training logic function (MTLF).

[0168] Example 52: A computer-readable medium comprising program instructions that, when executed by a means for network functions, cause the means to perform at least the following: - Obtain longitudinal federated learning VFL participant information representing at least one VFL participant, associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism; and - Determine the preferred security mechanism that has been specified; and - If a preferred security mechanism has already been specified: - Obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism; and - Prepare the VFL process using the first selected VFL participants and the preferred security mechanism for training a machine learning model.

[0169] Example 53: The computer-readable medium according to Example 52 further includes program instructions that, when executed by the device, cause the device to perform: - If a preferred security mechanism has not yet been specified: - Identify a common security mechanism supported by the maximum number of VFL participants among the at least one VFL participant; - Obtain or execute a second selection of a VFL participant among the at least one VFL participant, wherein the second-selected VFL participant supports the common security mechanism; and - Prepare the VFL process using the second selected VFL participant and the common security mechanism for training a machine learning model.

[0170] Example 54: The computer-readable medium according to Example 52 or 53 further includes program instructions that, when executed by the device, cause the device to perform: - Register the at least one VFL participant, wherein registering the corresponding VFL participant includes providing VFL participant identification information associated with the corresponding security capability information, wherein the VFL participant identification information indicates the corresponding VFL participant.

[0171] Example 55: The computer-readable medium according to any one of Examples 52 to 54 further includes program instructions that, when executed by the device, cause the device to perform: - Request to initiate the VFL process.

[0172] Implementation 56: A computer-readable medium according to any one of embodiments 52 to 55, wherein obtaining the VFL participant information includes: - Request and / or perform discovery of VFL participants.

[0173] Example 57: According to the computer-readable medium of Example 56, obtaining the VFL participant information includes: - As a result of the discovery of VFL participants, obtain a list representing the at least one VFL participant.

[0174] Example 58: According to the computer-readable medium of Example 57, obtaining the VFL participant information includes: - For at least one VFL participant in the list, obtain security capability information associated with the corresponding VFL participant, wherein the corresponding security capability information is obtained together with the list.

[0175] Implementation Method 59: The computer-readable medium according to Embodiment 57 or 58, wherein obtaining VFL participant information includes: - For at least one VFL participant in the list whose associated security capability information has not yet been obtained along with the list, request the corresponding security capability information from the relevant VFL participant; and / or - Obtain the corresponding security capability information from the corresponding VFL participants.

[0176] Example 60: The computer-readable medium according to any one of Examples 52 to 59 further includes program instructions that, when executed by the device, cause the device to perform: - Initiate a security negotiation for the at least one VFL participant represented by the VFL participant information.

[0177] Implementation 61: According to any one of embodiments 52 to 60, the computer-readable medium, wherein preparing the VFL process for training a machine learning model using the first selected VFL participants and the preferred security mechanism includes at least one of the following: - Finalize the selection of VFL participants; - Instantiate the VFL session identifier; - Configure the use of the preferred security mechanisms.

[0178] Example 62: The computer-readable medium according to Example 53, wherein preparing the VFL process for training a machine learning model using the second-selected VFL participants and the common security mechanism includes at least one of the following: - Finalize the selection of VFL participants; - Instantiate the VFL session identifier; - Configure the use of the aforementioned common security mechanism.

[0179] Example 63: According to the computer-readable medium of Example 55, requesting to initiate the VFL procedure includes: specifying the preferred security mechanism.

[0180] Implementation method 64: A computer-readable medium according to any one of embodiments 52 to 63, wherein the at least one security mechanism is at least one of the following: - Security mechanisms based on encryption; - A security mechanism based on privacy protection.

[0181] Example 65: A computer-readable medium according to any one of Examples 52 to 64, wherein at least one parameter of the at least one security mechanism varies according to at least one of the following: - The identifier of the machine learning model to be trained using the VFL process; - VFL participants involved in the VFL process.

[0182] Example 66: A computer-readable medium according to any one of Examples 52 to 65, wherein the computer-readable medium is a non-volatile computer-readable medium.

[0183] Any connections presented in the described embodiments are to be understood in a manner that the components involved are operatively coupled. Therefore, connections can be direct or indirect, have any number of intermediate elements or combinations of intermediate elements, and may exist solely as functional relationships between components.

[0184] Furthermore, as used herein, the term "circuit" refers to any of the following: (a) Hardware-only circuit implementation (e.g., implementation using only analog and / or digital circuits); (b) A combination of circuitry and software (and / or firmware), for example: (i) A combination of processors, or (ii) (one or more) processors / software (including portions of (one or more) digital signal processors), software, and (one or more) memories, which work together to enable a device, such as a mobile phone, to perform various functions); and (c) Circuits, such as one or more microprocessors or a portion thereof, that require software or firmware to operate, even if the software or firmware is not physically present.

[0185] This definition of "circuit" applies to all uses of the term herein, including in any claim. As a further example, as used in this application, the term "circuit" also covers implementations of a processor (or processors), or a portion thereof, and its accompanying software and / or firmware. The term "circuit" also covers, for example, baseband integrated circuits or application processor integrated circuits for mobile phones.

[0186] Any processor mentioned in this article, in particular but not limited to Figure 5The processor 501 can be any suitable type of processor. Any processor may include (but is not limited to): one or more microprocessors, one or more processors with an accompanying digital signal processor, one or more processors without an accompanying digital signal processor, one or more application-specific computer chips, one or more field-programmable gate arrays (FPGAs), one or more controllers, one or more application-specific integrated circuits (ASICs), or one or more computers. The associated architecture / hardware has been programmed in a manner that performs the described functions.

[0187] Furthermore, any action or step described or illustrated herein may be implemented using executable instructions in a general-purpose or special-purpose processor and stored on a computer-readable storage medium (e.g., a disk, memory, etc.) for execution by such a processor. The reference to "computer-readable storage medium" should be understood to encompass special-purpose circuitry such as FPGAs, ASICs, signal processing devices, and other devices.

[0188] As used herein, “at least one of the following: ” and “at least one of a list of two or more elements>” and similar wording, wherein a list of two or more elements is connected by “and” or “or”, indicates at least any one of the elements, or at least any two or more of the elements, or at least all of the elements.

[0189] The wording “A, or B, or C, or a combination thereof” or “at least one of A, B and C” can be understood as a non-exhaustive expression and includes at least the following: (i) A, or (ii) B, or (iii) C, or (iv) A and B, or (v) A and C, or (vi) B and C, or (vii) A and B and C.

[0190] It should be understood that the embodiments disclosed herein are merely exemplary, and any feature presented for a particular exemplary embodiment may be used alone in any aspect of this disclosure, or in combination with any feature presented for the same or another particular exemplary embodiment, and / or in combination with any other feature not mentioned. It will also be understood that any feature presented for example embodiments in a particular category may also be used in a corresponding manner in any other category of example embodiments.

[0191] Abbreviations 3GPP Third Generation Partnership Project 5GC 5G core 5GS 5G system AF application function AI (Artificial Intelligence) AMF Access and Mobility Management Functions AnLF Analysis Logic Function AUSF Authentication Server Functionality DN Data Network gNB gNodeB ML Machine Learning MTLF model training logic function NEF Network Open Functions NF Network Functions NR New Radio NRF Network Repository Functionality NWDAF Network Data Analysis Function PCF policy control function RAN (Radio Access Network) SMF Session Management Function UDM Unified Data Management UE User Equipment UPF User Face Functions VFL Vertical Federated Learning

Claims

1. A method comprising: - Obtain longitudinal federated learning VFL participant information representing at least one VFL participant associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism. - Determine the preferred security mechanism that has been specified; and - If a preferred security mechanism has already been specified: - Obtain or execute a first selection of a VFL participant among the at least one VFL participant, wherein the first selected VFL participant supports the preferred security mechanism; as well as - Prepare the VFL process using the first selected VFL participants and the preferred security mechanism for training a machine learning model.

2. The method according to claim 1, further comprising: - If a preferred security mechanism has not yet been specified: - Identify a common security mechanism supported by the maximum number of VFL participants among the at least one VFL participant; - Obtain or execute a second selection of a VFL participant among the at least one VFL participant, wherein the second selected VFL participant supports the common security mechanism; as well as - Prepare the VFL process using the second selected VFL participant and the common security mechanism for training a machine learning model.

3. The method according to claim 1 or 2, further comprising: - Register the at least one VFL participant, wherein registering the corresponding VFL participant includes providing VFL participant identification information associated with the corresponding security capability information, wherein the VFL participant identification information indicates the corresponding VFL participant.

4. The method according to any one of claims 1 to 3, further comprising: - Request to initiate the VFL process.

5. The method according to any one of claims 1 to 4, wherein, Obtaining the VFL participant information includes: - Request and / or perform discovery of VFL participants.

6. The method according to claim 5, wherein, Obtaining the VFL participant information includes: - As a result of the discovery of VFL participants, obtain a list representing the at least one VFL participant.

7. The method according to claim 6, wherein, Obtaining the VFL participant information includes: - For at least one VFL participant in the list, obtain security capability information associated with the corresponding VFL participant, wherein the corresponding security capability information is obtained together with the list.

8. The method according to claim 6 or 7, wherein, Obtaining the VFL participant information includes: - For at least one VFL participant in the list whose associated security capability information has not yet been obtained along with the list, request the corresponding security capability information from the relevant VFL participant; and / or - Obtain the corresponding security capability information from the corresponding VFL participants.

9. The method according to any one of claims 1 to 8, further comprising: - Initiate a security negotiation for the at least one VFL participant represented by the VFL participant information.

10. The method according to any one of claims 1 to 9, wherein, Preparing the VFL process for training a machine learning model using the first-selected VFL participants and the preferred security mechanism includes at least one of the following: - Finalize the selection of VFL participants; - Instantiate the VFL session identifier; - Configure the use of the preferred security mechanisms.

11. The method according to claim 2, wherein, Using the second-selected VFL participants and the common security mechanism to prepare the VFL process for training a machine learning model includes at least one of the following: - Finalize the selection of VFL participants; - Instantiate the VFL session identifier; - Configure the use of the aforementioned common security mechanism.

12. The method according to claim 4, wherein, The request to initiate the VFL process includes specifying the preferred security mechanism.

13. The method according to any one of claims 1 to 12, wherein, The at least one security mechanism is at least one of the following: - Security mechanisms based on encryption; - A security mechanism based on privacy protection.

14. The method according to any one of claims 1 to 13, wherein, At least one parameter of the at least one security mechanism varies according to at least one of the following: - Identifier of the machine learning model that will be trained using the VFL process; - VFL participants involved in the VFL process.

15. The method according to any one of claims 1 to 14, wherein, The method is performed by at least one network function.

16. An apparatus for network functions, the apparatus comprising at least one processor and at least one memory, the at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus to perform at least the following: - Obtain longitudinal federated learning VFL participant information representing at least one VFL participant associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism. - Determine the preferred security mechanism that has been specified; and - If a preferred security mechanism has already been specified: - Obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism; and - Prepare the VFL process using the first selected VFL participants and the preferred security mechanism for training a machine learning model.

17. The apparatus according to claim 16, wherein, The at least one memory also stores instructions that, when executed by the at least one processor, cause the device to perform the method according to any one of claims 2 to 15.

18. The apparatus according to claim 16 or 17, wherein, The network functionality includes the device or is included in the device.

19. The apparatus according to any one of claims 16 to 18, wherein, The network function is the VLF coordinator.

20. The apparatus according to any one of claims 16 to 19, wherein, The network function is the Network Data Analysis Function (NWDAF), which includes the Model Training Logic Function (MTLF).

21. An apparatus for network functions, the apparatus comprising: - A component for obtaining longitudinal federated learning VFL participant information representing at least one VFL participant associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism. - Components used to determine the preferred security mechanism; and - If a preferred security mechanism has already been specified: - Components for obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism; and - A component for preparing a VFL process for training a machine learning model using the first selected VFL participants and the preferred security mechanism.

22. The apparatus of claim 21, further comprising a component for performing the method of any one of claims 2 to 15.

23. The apparatus according to claim 21 or 22, wherein, The network functionality includes the device or is included in the device.

24. The apparatus according to any one of claims 21 to 23, wherein, The network function is the VLF coordinator.

25. The apparatus according to any one of claims 21 to 24, wherein, The network function is the Network Data Analysis Function (NWDAF), which includes the Model Training Logic Function (MTLF).

26. A non-volatile computer-readable medium comprising program instructions that, when executed by a means for network functions, cause the means to perform at least the following: - Obtain longitudinal federated learning VFL participant information representing at least one VFL participant associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism. - Determine the preferred security mechanism that has been specified; and - If a preferred security mechanism has already been specified: - Obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism; and - Prepare the VFL process using the first selected VFL participants and the preferred security mechanism for training a machine learning model.

27. The non-transitory computer-readable medium of claim 26, further comprising program instructions that, when executed by the means, cause the means to perform the method of any one of claims 2 to 15.

28. A computer-readable medium comprising program instructions that, when executed by a means for network functions, cause the means to perform at least the following: - Obtain longitudinal federated learning VFL participant information representing at least one VFL participant associated with corresponding security capability information, wherein the security capability information indicates that the corresponding VFL participant supports at least one security mechanism. - Determine the preferred security mechanism that has been specified; and - If a preferred security mechanism has already been specified: - Obtaining or performing a first selection of a VFL participant among the at least one VFL participant, wherein the first-selected VFL participant supports the preferred security mechanism; and - use The first selected VFL participants and the preferred security mechanism are used to prepare the VFL process for training a machine learning model.

29. The computer-readable medium of claim 28, further comprising program instructions that, when executed by the apparatus, cause the apparatus to perform the method of any one of claims 2 to 15.

Images