VFL

Vertical Federated Learning

Other →
Introduced in Rel-19 Also in: Services, Management

VFL is a privacy-preserving distributed machine learning framework where multiple parties collaboratively train a model using different feature sets from the same set of users without exposing raw data.

Category
Other
Introduced
Rel-19
Where
Core Network › 5G Core
Also touches
2 segments
Specifications
14 specs
VFL Description Purpose Related Detected Changes Specifications

Description

Vertical Federated Learning (VFL) is a specialized distributed machine learning paradigm standardized by 3GPP to enable collaborative AI model training across different organizations or network domains without centralizing raw, sensitive data. In contrast to horizontal federated learning where participants share the same feature space but different user samples, VFL is characterized by participants holding different features or attributes for the same set of overlapping user IDs. A typical scenario involves a mobile network operator holding radio access network (RAN) measurement data and an Over-The-Top (OTT) service provider holding application-layer quality data for the same subscribers. VFL allows these parties to jointly train a more comprehensive and accurate model—for instance, for predicting user experience—while keeping their respective datasets private and on-premises.

The technical operation of VFL involves a structured protocol with roles such as the guest party, host party(s), and potentially a coordinator. The process begins with privacy-preserving entity alignment, where the participating parties use cryptographic techniques like Private Set Intersection (PSI) to securely identify their common users without revealing non-overlapping IDs. Once the aligned user set is established, the collaborative training commences. A common architecture splits the model into a bottom model and a top model. Each party trains its own bottom model on its local feature set. The outputs (embeddings or intermediate results) from these bottom models are then securely aggregated, often via homomorphic encryption or secure multi-party computation (MPC), to compute the loss and gradients for the top model. These gradients are distributed back to each party to update their respective bottom models, all without any party seeing the raw features or labels of another.

Key components in the 3GPP VFL framework include the Network Data Analytics Function (NWDAF) which can act as a participant or coordinator, standardized interfaces for federated learning orchestration (e.g., Naf_FederatedLearning), and security protocols for secure aggregation and model exchange. The architecture is designed to integrate with the 5G Service-Based Architecture (SBA), allowing network functions like the AMF, SMF, and PCF to contribute data to federated learning processes. VFL's role is to unlock the value of partitioned data silos within the telecom ecosystem, enabling advanced AI/ML use cases such as joint network-service optimization, churn prediction, and personalized QoS management, while strictly adhering to data privacy regulations like GDPR.

Purpose & Motivation

VFL was introduced to address the critical challenge of data silos and privacy constraints that hinder the development of advanced AI-driven network and service management. In the telecom industry, valuable data is fragmented across operators, vendors, and service providers. For example, an operator has detailed network performance data, while a content provider has rich application behavior data. Individually, these datasets provide a limited view; combined, they could power highly accurate predictive models. However, legal, regulatory, and competitive barriers prevent the sharing or centralization of this raw data. Traditional methods of data pooling or model training on centralized datasets are thus infeasible, limiting the potential of AI in 5G and beyond.

The standardization of VFL in 3GPP Release 19 was motivated by the need to foster a trusted data collaboration ecosystem for 6G preparation and advanced 5G-Advanced networks. It solves the problem by providing a standardized, secure framework for collaborative learning that preserves data sovereignty. This enables participants to benefit from the combined predictive power of distributed feature sets while providing technical and procedural guarantees that raw data never leaves its owner's control. VFL unlocks new business models and operational efficiencies, such as co-developing churn prediction models with banking partners or optimizing video streaming jointly with content delivery networks, all within a privacy-by-design framework that builds trust among stakeholders.

Detected Changes Across Releases

from 3GPP Change Requests

Specific changes extracted from the „Change history“ tables of 3GPP specifications (280 CRs across 6 releases). Complements the general historical overview above with the evidence-based evolution of this function.

Rel-15 1 change

In Release 15, the introduction of the VFL function was focused on foundational corrections and clarifications. Specifically, this involved correcting the NWDAF resource to ensure proper support for vertical federated learning operations. This established the necessary groundwork for VFL's data analytics and model training capabilities within the network.

Rel-16 22 changes

In Release 16, the enhancements for the NWDAF, which supports the VFL function, introduced new analytics consumers including the UDM and OAM, and clarified its role in charging functions. The release added specific data handling capabilities, such as collecting MDT/SON parameters and defining a maxObjectNbr attribute for analytics services. Furthermore, it provided clarifications and corrections on key procedures, including UE mobility analytics, abnormal behaviour analysis, and the assertion of probability for analytics events.

  • NWDAF Discovery and Selection TS 29.510CR0148
  • Services invoked by NWDAF TS 29.510CR0239
  • Update the Analytics information provided by NWDAF TS 23.288CR0015
  • Optionality of data to be collected by NWDAF TS 23.288CR0040
  • Probability assertion clarification on NWDAF services description TS 23.288CR0045
  • Clarification on NWDAF-assisted expected UE behavioural analytics TS 23.288CR0087

+ 16 more changes

Rel-17 66 changes

In Release 17, the enhancements for Vertical Federated Learning (VFL) introduced new procedures and architectural support for coordinating multiple NWDAF instances. Key additions included the procedure for multiple NWDAF analytics aggregation, time coordination across instances, and NWDAF discovery and selection based on ML Model information. Furthermore, the release specified service operations for analytics aggregation, NWDAF reselection procedures, and the ability to register NWDAF instances into the UDM.

  • NWDAF decomposition TS 23.288CR0205
  • NWDAF - Data repository function TS 23.288CR0206
  • Procedure for Multiple NWDAF Analytics aggregation TS 23.288CR0207
  • Procedure for time coordination across multiple NWDAF instances TS 23.288CR0208
  • Implementation of Enhancements on Event Exposure used by NWDAF in TS23.288 TS 23.288CR0216
  • ML Model sharing between NWDAF instances TS 23.288CR0218

+ 60 more changes

Rel-18 89 changes

In Release 18, the VFL function introduced Federated Learning among Multiple NWDAFs, establishing procedures for Model Information Exchange and the Maintenance of the Federated Learning Process within the 5G Core. It specifically added capabilities for NWDAF discovery and selection for an NWDAF supporting MTLF with FL capability, and enhanced analytics like Service Experience and WLAN performance analytics to assist federated learning operation and member selection.

  • Alignment for UPF event exposure service to NWDAF via the SMF in TS 23.288 TS 23.288CR0551
  • Enhancement on OSE for NWDAF assisting PCF in making URSP decisions TS 23.288CR0569
  • NWDAF updates to assist resource monitoring of AI/ML-based services TS 23.288CR0575
  • Federated Learning among Multiple NWDAFs in TS 23.288 TS 23.288CR0582
  • NWDAF-assisted application detection in TS 23.288 TS 23.288CR0584
  • Adding Accuracy Checking Capability to NWDAF Architecture TS 23.288CR0564

+ 83 more changes

Rel-19 100 changes

In Release 19, the Vertical Federated Learning (VFL) function introduced new, standardized procedures for both training and inference between network functions like the NWDAF and AFs. It specifically defined registration and discovery procedures for VFL participants and added support for client intermediate results sharing between VFL clients during model training and inference. The release also included enhancements for VFL model accuracy monitoring and refinements to resolve open issues in the general VFL procedures.

  • General inference procedure for vertical federated learning TS 23.288CR1126
  • Registration and Discovery procedure for Vertical Federated Learning among NWDAF(s) and/or AF(s) with NWDAF as the VFL server TS 23.288CR1171
  • High level feature description for VFL TS 23.288CR1185
  • Refinements for VFL feature TS 23.288CR1198
  • KI#2 - Update of VFL training and inference TS 23.288CR1246
  • Update the general inference procedure for vertical federated learning to resolve ENs TS 23.288CR1208

+ 94 more changes

Rel-20 2 changes

In Release 20, the VFL function introduced new capabilities to enhance federated learning for vertical industries. Specifically, it added support for sample alignment procedures to enable VAL servers to coordinate training data. Furthermore, the release extended split learning functionality by defining mechanisms to support the use of relay nodes for distributing machine learning model partitions.

  • Sample Alignment Enablement for VAL Servers in VFL TS 23.482CR0062
  • Support for ML model split learning using relays TS 23.482CR0081

Explore further

Broader topics and technologies where VFL plays a role.

Defining Specifications

3GPP specifications that define or reference VFL, with the latest known release. Sourced from the 3GPP document catalog — see methodology.

SpecificationTitleRelease
TR 21.905 vj00 3GPP Technical Terms and Definitions Rel-19
TS 23.288 vk00 5GS Architecture Enhancements for Data Analytics Rel-20
TS 23.482 vk00 AIML Enablement Service Architecture Rel-20
TS 23.700 vk00 XR Services Application Enablement Layer Rel-20
TS 24.560 vj00 AIML Enablement (AIMLE) Services Stage 3 Protocol Rel-19
TS 28.105 vj30 AI/ML Management for 5GS Rel-19
TS 28.858 vj00 AI/ML Management Phase 2 Study Rel-19
TS 29.510 vj50 NRF Service Based Interface Protocol Rel-19
TS 29.520 vj40 5G Network Data Analytics Services Stage 3 Rel-19
TS 29.530 vj00 AF AI/ML Services Stage 3 Protocol Rel-19
TS 29.552 vj40 5G Network Data Analytics Signalling Flows Rel-19
TS 29.591 vj40 5G NEF Southbound Services Stage 3 Rel-19
TS 33.501 vk00 5G Security Architecture and Procedures Rel-20
TS 33.784 vj00 Security aspects of AI/ML in core network Rel-19