Neural Network Structure for Federated Learning Data Protection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In federated learning processes, there is a risk of data leakage where the original feature information of a passive party can be reversely deduced from the target feature information transmitted to an active party, compromising data security due to high coupling between the original and target feature information.
Innovation Solution
A neural network structure is trained to minimize the coupling degree between original and target feature information by processing the original feature information using a target feature processing network structure, reducing the possibility of data leakage and enhancing data security.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If original feature information is transmitted to the active party for joint training, then model training collaboration is enabled, but data leakage risk increases due to high coupling between original and target feature information
Solution Approach 1:
The patent introduces a target feature processing network structure as an intermediary between the original feature information and the active party. This intermediary processes the original feature information to generate target feature information with reduced coupling, thereby enabling model training collaboration while mitigating data leakage risk through the protective transformation layer
2Reliability
If a neural network structure is trained to minimize coupling degree, then data security is enhanced, but training complexity and computational resources increase
Solution Approach 1:
The patent employs parameter changes by introducing a learnable transformation parameter matrix W in the target feature processing network structure. The coupling reduction is achieved by optimizing this parameter matrix during training to minimize the coupling degree between original and target feature information, thereby enhancing data security through parameter optimization rather than structural complexity
Data Source
AI summary
The present disclosure relates to a data protection method, a training method and apparatus for a network structure, a medium, and a device. The data protection method includes: obtaining original feature information of a target batch of reference samples for a passive party of a joint training model; and processing the original feature information by means of a target feature processing network structure to obtain target feature information corresponding to the original feature information. A neural network structure is trained by at least aiming at minimizing a coupling degree of between original training feature information and target training feature information of training samples for the passive party to obtain the target feature processing network structure. The target training feature information is feature information corresponding to the original training feature information that is outputted from the neural network structure using the original training feature information as an input.