Secure Federated Model Integration Across Heterogeneous Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face difficulties in building a network that connects multiple organizations with different security approaches, making it challenging to construct a global AI model from local models across these organizations.
Innovation Solution
A learning apparatus and system that establish secure communication with information terminals in each organization's network, acquire local models using secure protocols, and integrate them to form a global model, utilizing techniques like VPNs and secure computation to maintain confidentiality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If networks of multiple organizations are connected to build a global model, then model performance is improved, but security compatibility and network stability deteriorate due to different security approaches and non-constant connections
Solution Approach 1:
The patent introduces a learning apparatus as an intermediary component that mediates between multiple organization networks with different security approaches. This intermediary establishes secure communication channels, acquires local models from various organizations, and integrates them into a global model, thereby resolving the security compatibility issue while enabling model performance improvement
Solution Approach 2:
The patent segments the global model construction process into independent local model training phases within each organization's network, followed by centralized integration. This segmentation allows each organization to maintain its own security protocols independently while still contributing to the overall global model, thus maintaining security compatibility across diverse networks
2Reliability
If secure communication protocols are implemented to protect data confidentiality, then security is improved, but communication complexity and system overhead increase
Solution Approach 1:
The learning apparatus implements a universal secure communication framework that handles multiple security requirements through standardized protocols. By creating a multi-functional communication system that can accommodate different security approaches from various organizations, the patent reduces overall system complexity while maintaining data confidentiality
3Measurement precision
If local models are collected from multiple organizations, then global model accuracy is improved, but data collection difficulty and network connection requirements increase
Solution Approach 1:
The patent enables organizations to self-serve by training local models within their own networks using their own data sets. The learning apparatus then automatically collects these locally-trained models through secure communication channels, eliminating the need for manual data transfer and reducing network connection complexities while maintaining high global model accuracy
Data Source
AI summary
A learning apparatus includes: a communication establishment unit configured to establish secure communication with an information terminal arranged in a network of each one of organizations; an acquisition unit configured to acquire local models which have learned a data set for each of the organizations from a corresponding one of the information terminals using the secure communication; and an integration unit configured to integrate the plurality of local models that have been acquired.