Federated Learning Model Training with Encrypted Data Partitioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing frameworks fail to maintain privacy of private data during federated training and inference, especially when data is both vertically and horizontally partitioned, leading to potential data leakage and compliance issues in regulated fields like healthcare and finance.
Innovation Solution
A system employing a random decision tree-based approach for training inferential models using both horizontally and vertically partitioned data, ensuring that first-party private data is not directly shared with other parties, combined with differential privacy mechanisms to protect data privacy during inference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If federated training is employed between two or more entities to analyze partitioned data, then data analytics capability is improved, but privacy of private data is compromised
Solution Approach 1:
The patent introduces secure computation protocols and encryption mechanisms as intermediaries between participating entities. These intermediaries enable federated training by allowing computations on encrypted data without revealing the underlying private information, thus resolving the contradiction between analytics capability and privacy protection.
Solution Approach 2:
The patent segments data into different partitioning schemes (horizontal and vertical) and processes them through separate privacy-preserving mechanisms. By dividing the data processing into isolated segments that never directly expose private information, the system maintains both analytics capability and privacy.
2Ease of manufacture
If existing frameworks are used to analyze vertically partitioned data or horizontally partitioned data separately, then implementation simplicity is improved, but privacy protection is insufficient
Solution Approach 1:
The patent merges horizontal and vertical data partitioning frameworks into a unified federated training system. This combination allows the system to leverage the simplicity of existing separate frameworks while adding layered privacy protection mechanisms that address the insufficiencies of individual approaches.
3Measurement precision
If private data is directly shared among parties for federated training, then model training accuracy is improved, but data leakage risk increases
Solution Approach 1:
The patent changes the state of data from plaintext to encrypted form during federated training. By transforming data parameters (encryption levels, representation formats), the system enables accurate model training on encrypted data without exposing the underlying information, thus preventing data leakage while maintaining training accuracy.
Data Source
AI summary
Systems, devices, computer program products and/or computer-implemented methods of use provided herein relate to federated training and inferencing. A system can comprise a memory that stores computer executable components, and a processor that executes the computer executable components stored in the memory, wherein the computer executable components can comprise a modeling component that trains an inferential model using data from a plurality of parties and comprising horizontally partitioned data and vertically partitioned data, wherein the modeling component employs a random decision tree comprising the data to train the inferential model, and an inference component that responds to a query, employing the inferential model, by generating an inference, wherein first party private data, of the data, originating from a first passive party of the plurality of parties, is not directly shared with other passive parties of the plurality of parties to generate the inference.