Federated Learning Local Group Formation via Trust Value Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Federated learning systems face challenges in maintaining learning accuracy, stability, and security due to data scarcity, resource constraints, and malicious behavior among local clients within a local group.
Innovation Solution
A federated learning system and method that form local groups based on trust value evaluations, using a central server to request feature set information from nodes, designate a temporary master node, and generate local groups while excluding inappropriate nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If local groups are formed for federated learning, then data sharing and collaboration are improved, but learning accuracy and stability deteriorate due to insufficient resources and malicious behavior
Solution Approach 1:
The patent introduces a trust value evaluation mechanism as an intermediary between nodes in local groups. The central server calculates trust values based on multiple features (participation frequency, data quality, resource capability) and uses these trust values to identify and exclude malicious nodes. This mediator mechanism enables reliable data sharing while maintaining learning accuracy by filtering out harmful elements in the federated learning process.
2Adaptability or versatility
If nodes with insufficient resources are included in local groups, then group diversity and data coverage are improved, but system performance and stability deteriorate
Solution Approach 1:
The patent changes the parameter of node selection from simple inclusion to trust-value-based selection. The central server evaluates multiple parameters including participation frequency, data quality, and resource capability to calculate comprehensive trust values. Nodes are dynamically selected or excluded based on these trust values, allowing the system to maintain diverse groups while ensuring sufficient resource capability and performance stability.
3Device complexity
If malicious nodes are allowed to participate, then system complexity is reduced, but security risks and reliability deteriorate
Solution Approach 1:
The patent implements preliminary action by evaluating trust values before nodes participate in federated learning. The central server calculates trust values based on historical behavior, data quality, and resource capability before allowing nodes to join local groups. This preliminary evaluation prevents malicious nodes from participating, eliminating security risks before they can affect the system while maintaining relatively simple system architecture.
4Measurement precision
If centralized learning is performed, then learning accuracy is maintained, but data sharing and collaboration capabilities deteriorate
Solution Approach 1:
The patent applies segmentation by dividing the centralized learning system into multiple local groups with trusted nodes. Each local group performs federated learning independently with guaranteed accuracy through trust value evaluation, while still enabling data sharing and collaboration across groups. This segmentation maintains learning accuracy similar to centralized learning while restoring data sharing capabilities that were lost in purely decentralized approaches.
Data Source
AI summary
A federated learning system comprises at least one central server and a plurality of local groups, wherein each of the plurality of local groups comprises one master node and a plurality of nodes, wherein the plurality of local groups are formed by the central server using information of a feature set (C) for federated learning.