VFL Coordinator Authorization and Privacy-Preserving Sample Alignment

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing Vertical Federated Learning (VFL) environments face privacy breaches during the sample alignment process, with sensitive user equipment (UE) identifiers potentially exposed between participants, and there is a lack of robust authorization mechanisms for selecting participants across different domains in 3GPP networks.

Innovation Solution

Implement secure ID mapping and encryption protocols to protect UE identifiers, using Network Exposure Function (NEF) encryption and integrity protection, and establish a robust authorization framework with secure token-based mechanisms and local policies to ensure authorized participation in VFL processes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If sample alignment process is implemented in VFL, then participants can share data for model training, but UE identifiers are exposed between participants causing privacy breaches

Engineering Contradiction:
Improvemodel training efficiencyVSAvoidprivacy breach
Core Design Contradiction:
ProductivityVSObject-affected harmful factors

Solution Approach 1:

The patent introduces a VFL Coordinator as an intermediary component that manages the sample alignment process. The coordinator receives sample alignment requests from participants, performs the alignment operation using encrypted identifiers, and returns results without exposing raw UE identifiers to participants. This mediator architecture enables collaborative model training while protecting participant privacy through centralized controlled access.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent extracts and separates the identifier mapping function from the participant systems and places it within the VFL Coordinator. By taking out the sensitive identifier handling from participant control, the system prevents participants from directly accessing or exposing UE identifiers while still enabling the necessary sample alignment functionality for model training.

Inventive Principle:
Principle #2Taking out (Extraction)

2Reliability

If authorization mechanisms are implemented across different domains in 3GPP networks, then secure participant selection is achieved, but system complexity increases

Engineering Contradiction:
Improveauthorization securityVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The VFL Coordinator is designed as a universal component that handles multiple functions: sample alignment, authorization verification, token management, and participant coordination. By consolidating these diverse functions into a single multi-functional entity, the system achieves robust cross-domain authorization and secure participant selection without proportionally increasing overall system complexity.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent implements preliminary authorization actions where the VFL Coordinator verifies participant credentials and issues authorization tokens before participants can access the sample alignment process. This preliminary security check prevents unauthorized access and ensures only authenticated participants can join VFL groups, establishing security upfront rather than requiring complex continuous verification mechanisms.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4687327A1Advanced privacy protection mechanisms in and authorization mechanisms for vertical federated learning (VFL)
Publication Date: 2026.02.04 INTEL CORP
  • EP4687327A1 patent drawingFigure 1
  • EP4687327A1 patent drawingFigure 2~3
  • EP4687327A1 patent drawingFigure 4

AI summary

A non-transitory computer-readable medium including instructions which, when executed by a processor of a network repository function, NRF, cause the processor to: receive a discovery request from a vertical federated learning, VFL, server to identify eligible VFL clients based on registered network function, NF, profiles; match the discovery request against the registered NF profiles; respond with eligible VFL client identifiers; and issue a secure token to the VFL server after verifying interoperability indicators and analytics capabilities of the VFL clients based on the registered NF profiles.