Federated Learning Confidentiality Using MPC and Trusted Execution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional federated learning systems fail to ensure the confidentiality of local training data, making them susceptible to confidentiality attacks such as membership inference, attribute inference, and reconstruction attacks, and are vulnerable to malicious aggregators and orchestrators.
Innovation Solution
Implementing secure multi-party computation (MPC) protocols to encrypt and share machine learning model updates as secret shares, using identifiers instead of the actual models, and performing operations within trusted execution environments (TEE) to ensure confidentiality and integrity of the training process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional federated learning systems are used, then the system structure is simple and easy to implement, but data privacy is compromised and confidentiality attacks can succeed
Solution Approach 1:
The patent applies segmentation by dividing the machine learning model into secret shares using secure multi-party computation (MPC). Each participant holds only a portion of the model parameters, and no single entity can reconstruct the full model without cooperation from multiple parties. This segmentation prevents confidentiality attacks while maintaining the functional integrity of the federated learning system.
Solution Approach 2:
The patent introduces an intermediary mechanism through the use of MPC protocols and trusted execution environments (TEEs). These intermediaries enable secure communication and computation between participants without requiring direct trust between them. The intermediary layer ensures that data and model updates remain confidential while still allowing necessary computations to proceed.
2Reliability
If secure multi-party computation protocols are implemented, then data privacy and confidentiality are enhanced, but computational complexity and processing time increase
Solution Approach 1:
The patent applies preliminary action by pre-establishing MPC protocols and distributing secret sharing schemes before the actual training begins. The system setup phase involves configuring TEEs, establishing communication channels, and preparing the MPC framework in advance. This preliminary configuration enables faster execution during the training phase, as the heavy computational overhead is incurred once during setup rather than in each training iteration.
3Reliability
If trusted execution environments are used, then integrity and security of model updates are protected, but system complexity and resource requirements increase
Solution Approach 1:
The patent applies copying by creating virtual instances or copies of the execution environment within TEEs. Instead of requiring physical isolation and complex hardware configurations, the system creates software-based copies of the necessary computational environments. This allows multiple participants to have isolated execution spaces without requiring complex physical infrastructure, reducing overall system complexity while maintaining integrity.
Data Source
AI summary
A method comprising a computer-implemented method for federated learning for an owner of a machine learning model, a computer-implemented method for federated learning for an orchestrator, a computer-implemented method for federated learning for a training client, and/or a computer-implemented method for federated learning for an aggregator.