Federated Model Training With Adaptive Cloud-Edge Reporting

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing horizontal federated learning process is inflexible and inefficient due to strict model aggregation by edge servers, which does not align with actual network deployment scenarios, leading to slow communication and inconsistent training durations.

Innovation Solution

A method to dynamically select the reporting path of model information based on client capability information, allowing clients to report directly to the cloud server or through an edge server, optimizing the training process by considering memory, computing power, and network bandwidth.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If data is preprocessed and stored in a database before training, then training efficiency is improved, but server resources are consumed and privacy security is compromised

Engineering Contradiction:
Improvetraining efficiencyVSAvoidprivacy security
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

A trusted execution environment (TEE) is introduced as an intermediary between the client and server. The TEE creates an isolated secure computing space where data preprocessing and model training occur without exposing sensitive data to the untrusted server environment, thus enabling efficient processing while maintaining privacy security

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system segments the computing environment into two distinct parts: a trusted execution environment for secure data processing and model training, and an untrusted server environment for resource provision. This segmentation allows the server to provide computational resources without having access to or control over the sensitive data being processed

Inventive Principle:
Principle #1Segmentation

2Power

If data is transmitted to server for processing, then computational power is utilized, but transmission overhead and security risks increase

Engineering Contradiction:
Improvecomputational powerVSAvoidtransmission overhead
Core Design Contradiction:
PowerVSLoss of energy

Solution Approach 1:

The trusted execution environment is nested within the server's computational infrastructure. This allows the system to leverage the server's powerful hardware resources while maintaining an isolated secure computing space, eliminating the need for separate data transmission and reducing transmission overhead

Inventive Principle:
Principle #7Nested doll (Nesting)

3Manufacturing precision

If model parameters are updated and stored centrally, then model optimization is achieved, but security vulnerabilities increase

Engineering Contradiction:
Improvemodel optimizationVSAvoidsecurity vulnerabilities
Core Design Contradiction:
Manufacturing precisionVSObject-affected harmful factors

Solution Approach 1:

The system extracts and processes model parameters within the trusted execution environment on the client side. Only encrypted or processed results are transmitted to the server for storage, removing sensitive intermediate parameters from the untrusted server environment and eliminating security vulnerabilities associated with central parameter storage

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4300876B1Model training method and apparatus
Publication Date: 2026.05.06 HUAWEI TECH CO LTD
  • EP4300876B1 patent drawingFigure 1~2
  • EP4300876B1 patent drawingFigure 3~4
  • EP4300876B1 patent drawingFigure 5~6

AI summary

This application provides a model training method and apparatus. The method includes: obtaining first capability information and second capability information of a client; determining, based on the first capability information and the second capability information, that the client reports model information to a cloud server or an edge server; and sending first indication information to the client. The first indication information indicates the client to report the model information to the cloud server or the edge server. According to the model training method and apparatus provided in this application, in a horizontal federated training method in a hybrid mode, a manner of reporting local model information by the client may be flexibly selected based on capability information of the client, so that horizontal federated training efficiency can be improved.