Federated Learning Model Aggregation Without Trusted Third Party
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Federated learning model training relies on a trusted third party, leading to privacy data leakage and security risks due to centralized storage of private keys, making it difficult to implement effectively.
Innovation Solution
A method for federated learning that allows multiple node devices to jointly calculate second-order gradient descent directions without relying on a third-party node, using fusion operators to determine model parameters and gradients, thereby enhancing security and reducing single-point centralized risks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a trusted third party is used to coordinate model training in federated learning, then model aggregation can be centralized and simplified, but security risks increase and private data leakage becomes possible
Solution Approach 1:
The patent extracts the trusted third party from the system by implementing a distributed consensus mechanism where node devices collectively perform model aggregation through cryptographic protocols, eliminating the security vulnerability while maintaining coordination functionality
Solution Approach 2:
The patent introduces cryptographic intermediaries (secret sharing schemes and secure multi-party computation protocols) that enable secure model aggregation without requiring a trusted third party, allowing nodes to coordinate while protecting private data
2Device complexity
If a trusted third party stores private keys centrally, then key management becomes simplified, but single-point security risks increase and the system becomes difficult to implement
Solution Approach 1:
The patent segments the private key storage and management function across multiple node devices using secret sharing schemes, where each node holds a share of the key material, eliminating the single-point security risk while distributing the complexity of key management
Solution Approach 2:
The patent combines key management functionality into the distributed consensus protocol itself, where model aggregation and secure key operations are performed collectively by nodes through cryptographic protocols, simplifying the overall system architecture without centralizing security risks
Data Source
AI summary
A model training method and apparatus for federated learning, a device and a storage medium are provided, which belong to the technical field of machining learning. The method includes: generating an ith scalar operator based on a (t-1)th round of training data and a tth round of training data (201); transmitting an ith fusion operator to a next node device based on the ith scalar operator (202); determining an ith second-order gradient descent direction of an ith sub-model based on an acquired second-order gradient scalar, an ith model parameter and an ithfirst-order gradient; and updating the ith sub-model based on the ith second-order gradient descent direction to obtain a model parameter of the ith sub-model during a (t+1)th round of iterative training.