Secret Sharing for Secure Model Parameter Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In the context of cooperative data processing among enterprises, there is a challenge in determining model parameters while protecting data privacy, as existing methods fail to ensure secure sharing of sensitive information without compromising data security.
Innovation Solution
The method involves secretly sharing a Hessian matrix, its inverse, and new model parameters between cooperation partners using secret sharing algorithms, allowing for collaborative model parameter determination without exposing individual data, leveraging the Newton's method for high convergence speed and reduced optimization iterations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If enterprises cooperatively train data processing models using distributed data, then model training effectiveness is improved, but data privacy security deteriorates
Solution Approach 1:
The patent segments the model training process into multiple iterations where only model parameters (not raw data) are shared between enterprises. Each enterprise keeps its data locally and contributes only to the collaborative optimization of model parameters through repeated iterations, thus maintaining data privacy while achieving effective model training.
Solution Approach 2:
The patent introduces model parameters as an intermediary that mediates the collaboration between enterprises. Instead of directly sharing sensitive data, enterprises interact through the model parameters which are updated iteratively and shared across parties, enabling cooperative training without exposing underlying data.
2Measurement precision
If model parameters are optimized multiple times using conventional methods, then model accuracy is improved, but computation time increases
Solution Approach 1:
The patent performs preliminary computation of the Hessian matrix and its inverse at the beginning of the training process. By pre-computing these components, the patent reduces the computational burden in subsequent optimization iterations, enabling faster convergence while maintaining model accuracy.
Solution Approach 2:
The patent changes the optimization approach by using the Newton method with pre-computed Hessian matrix inverse, rather than conventional gradient descent. This parameter change in the optimization algorithm significantly reduces the number of iterations needed for convergence, thereby reducing computation time while improving model accuracy.
Data Source
AI summary
This disclosure relates to determining model parameters using secret sharing. In some aspects, a first data party device obtains a first share of a Hessian matrix for a data processing model. The first data party device obtains, using secret sharing with the second data party device, a first share of a product of a random number matrix and the Hessian matrix. The first data party device, determines a first share of a first inverse matrix based on a second inverse matrix and the first share of the random number matrix. The first data party device determines the first inverse matrix, a first share of a product of the first inverse matrix and a gradient of a loss function of the data processing model, and a first share of a new model parameter for the data processing model.