Hierarchical Multiparty Homomorphic Encryption for Federated Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multiparty homomorphic encryption protocols face issues such as single points of failure, high time overheads, increased communication costs, and the need for system resets due to configuration updates, especially in federated learning scenarios involving geographically distributed devices like IoT devices.
Innovation Solution
A hierarchical system is implemented with abstracted devices forming clusters, allowing different multiparty homomorphic encryption protocols to be used based on device reliability, reducing communication costs and simplifying configuration updates by isolating clusters and using gateway devices for intermediate aggregation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If fully-multiparty homomorphic encryption (FMHE) is implemented to eliminate single points of failure, then security is improved, but the system becomes sensitive to contributor delays and requires system resets for configuration updates
Solution Approach 1:
The system segments contributors into multiple clusters, each running independent MHE protocols. This segmentation allows the system to avoid full system resets by isolating configuration changes to specific clusters, thereby reducing the time overhead associated with system-wide resets while maintaining security through distributed architecture.
Solution Approach 2:
The system dynamically selects which MHE protocol each cluster implements based on contributor reliability. More reliable devices run FMHE for higher security, while less reliable devices run PHE. This dynamic allocation optimizes the balance between security and time overhead by matching protocol complexity to device capability.
2Reliability
If threshold-multiparty homomorphic encryption (TMHE) is used to improve resilience to dropout, then reliability is improved, but communication costs and time overhead increase
Solution Approach 1:
By segmenting the system into clusters that independently run MHE protocols, communication is localized within clusters rather than requiring system-wide communication. This reduces overall communication costs while maintaining resilience through the threshold-based security model at the cluster level.
Solution Approach 2:
Different clusters can implement different MHE protocols based on their specific reliability requirements and communication constraints. This allows the system to optimize the balance between resilience and communication costs locally at each cluster rather than applying a uniform approach system-wide.
3Reliability
If a uniform MHE protocol is implemented across all devices, then security consistency is improved, but flexibility to accommodate varying device reliability is reduced
Solution Approach 1:
The system applies different MHE protocols to different clusters based on their specific reliability characteristics. This allows each cluster to have optimized security consistency appropriate to its devices, while the overall system maintains flexibility in protocol selection.
Solution Approach 2:
The gateway device serves multiple functions: it acts as a client device participating in cluster MHE operations, and simultaneously serves as a server for other client devices. This multi-functionality enables flexible protocol selection while maintaining security consistency through the gateway's coordination role.
4Adaptability or versatility
If configuration updates are required to add, remove or update contributors, then adaptability is improved, but the entire system must be reset causing loss of time
Solution Approach 1:
The system segments contributors into independent clusters, allowing configuration updates to be applied locally to affected clusters rather than requiring system-wide resets. This maintains adaptability for configuration changes while significantly reducing the time overhead by isolating updates to specific clusters only.
Solution Approach 2:
The gateway device acts as an intermediary that manages configuration updates within its cluster. When configuration changes are needed, the gateway coordinates the update locally without triggering system-wide resets, thereby maintaining adaptability while reducing time overhead through localized update management.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A hierarchical system configured to implement a secure multiparty computation protocol comprises: a plurality of client devices arranged in a plurality of C clusters of client devices, each cluster comprising: one or more client devices of the plurality of client devices; and a gateway device configured to communicate with the one or more client devices in the cluster, wherein, within each cluster, the one or more client devices are configured to implement a respective multiparty homomorphic encryption (MHE) protocol associated with that cluster; wherein: the plurality of C gateway devices form a gateway cluster of gateway devices, each gateway device of the plurality of C gateway devices configured to communicate with the other gateway devices; the gateway devices in the gateway cluster are configured to implement a gateway multiparty homomorphic encryption (MHE) protocol; and one or more gateway devices of the plurality of C gateway devices is configured to communicate with a central server.