Federated Learning Parameter Relay for Secure Client-Server Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing federated learning methods fail to ensure the security of model parameters transmitted between clients and a server, making them vulnerable to data leakage risks if stolen by attackers.
Innovation Solution
Implement a federated learning system where terminals send trained model parameters to a server through collaborative terminals with established relationships, using encryption noise to secure the transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If model parameters are transmitted directly from terminals to server in federated learning, then training efficiency is improved, but data security deteriorates as attackers can steal parameters and reconstruct local data
Solution Approach 1:
The patent introduces collaborative terminals as intermediaries between the original terminal and the server. The model parameters are first transmitted to collaborative terminals, which then forward them to the server. This intermediary mechanism prevents direct transmission while maintaining training efficiency, as the collaborative terminals act as secure relay points that prevent attackers from directly intercepting parameters from the original terminal.
Solution Approach 2:
The patent segments the parameter transmission path by involving multiple collaborative terminals instead of direct terminal-to-server communication. The parameters are divided and transmitted through different collaborative terminals, making it harder for attackers to reconstruct the complete parameter set and subsequently reconstruct local data.
2Measurement precision
If centralized learning mode is used to collect and analyze data, then machine learning performance is improved, but privacy protection deteriorates as users lose control over their data
Solution Approach 1:
The patent segments the data processing function by keeping raw data distributed at local terminals while only transmitting processed model parameters to the server. This segmentation allows the system to maintain high machine learning performance through centralized parameter aggregation while preserving user privacy control, as the raw sensitive data never leaves the local terminals.
Solution Approach 2:
The patent uses collaborative terminals as intermediaries that facilitate the federated learning process without requiring users to directly share their raw data. This intermediary mechanism enables machine learning performance improvement through collective learning while maintaining privacy control, as users only expose encrypted model parameters rather than their underlying sensitive data.
3Loss of information
If federated learning is used to keep data local, then privacy protection is improved, but data security deteriorates as model parameters remain vulnerable to theft and reconstruction attacks
Solution Approach 1:
The patent introduces collaborative terminals as secure intermediaries that receive and forward model parameters. This intermediary layer adds security by preventing direct access to parameters from the original terminal, making it more difficult for attackers to steal parameters and reconstruct local data while maintaining the privacy protection benefits of federated learning.
Solution Approach 2:
The patent segments the parameter transmission through multiple collaborative terminals, which dilutes the security risk. Even if one collaborative terminal is compromised, the complete parameter set cannot be easily reconstructed, thereby enhancing data security while maintaining privacy protection through local data retention.
Data Source
AI summary
The technology of this application relates to a training method that includes a first terminal obtaining a to-be-trained first machine learning model from the server. The first terminal is any one of a plurality of terminals. The first terminal trains the first machine learning model by using local data stored by the first terminal, to obtain trained model parameters. The first terminal determines, based on a collaboration relationship, a first collaborative terminal corresponding to the first terminal, and sends a part or all of the trained model parameters of the first terminal to the server by using the first collaborative terminal. The collaboration relationship is delivered by the server to the first terminal. The foregoing manner can improve security of data exchange between the server and the terminal.