Secure Multi-Party Extremum Search With Mean-Based Node Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing secure multi-party computation methods for determining an extremum face high computation complexity and communication overhead, especially in multi-party scenarios, making them inefficient for cross-network communication applications.
Innovation Solution
A method involving federated computation systems where node devices compare their data with progressively decreasing means, replacing data less than the mean with invalid data to converge on the extremum, reducing the need for encryption and decryption operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If oblivious transfer technology is used for multi-party data comparison, then data security is maintained, but communication overhead increases and efficiency decreases
Solution Approach 1:
The patent segments the multi-party comparison problem into multiple rounds of computation, where in each round only a subset of parties actively compare their data. This reduces the communication overhead from O(n²) to O(n) per round, while maintaining security through the secure comparison protocol. The segmentation of computation rounds allows the system to scale to more parties without quadratic communication costs.
Solution Approach 2:
The patent introduces a filtering mechanism that performs preliminary actions by identifying and eliminating parties with non-extremum values before the final comparison round. This preliminary filtering reduces the number of active participants in subsequent rounds, thereby reducing communication overhead while ensuring that the security-critical comparison rounds involve fewer parties.
2Reliability
If homomorphic encryption technology is used for value comparison, then data privacy is protected, but computation complexity increases
Solution Approach 1:
The patent applies local quality by using homomorphic encryption only for the critical comparison operations where privacy is most needed, rather than encrypting all data throughout the entire computation process. In non-critical phases such as data filtering and mean calculation, the system uses plaintext or less computationally intensive encryption methods, thereby reducing overall computation complexity while maintaining privacy where essential.
Solution Approach 2:
The patent changes the encryption parameter state by switching between different encryption modes based on the computation phase. During rounds requiring high privacy protection, full homomorphic encryption is applied; during intermediate rounds, the system transitions to lighter cryptographic primitives or plaintext processing, dynamically adjusting the security-computation tradeoff based on the specific requirements of each computation stage.
3Measurement precision
If multiple communication rounds are executed in multi-party scenarios, then comparison accuracy is improved, but communication overhead increases
Solution Approach 1:
The patent implements partial action by not requiring all n parties to participate in every comparison round. Instead, in each round, only a selected subset of parties (those likely to have extremum values based on previous rounds' results) actively participate in the comparison. This partial participation maintains comparison accuracy for the extremum detection while significantly reducing the communication overhead from O(n²) to O(n) per round.
Solution Approach 2:
The patent uses preliminary filtering in earlier rounds to identify and focus subsequent comparison rounds on a smaller subset of candidate parties. This preliminary action reduces the search space for the extremum, allowing the system to achieve the same comparison accuracy with fewer full-round communications, thereby reducing total communication overhead across all rounds.
Data Source
AI summary
Method for determining an extremum based on secure multi-party computation includes: acquiring a tth mean obtained in tth round federated computation; performing (t+1)th round federated computation based on valid node data of an ith node device in response to the valid node data of the ith node device being greater than the tth mean; performing the (t+1)th round federated computation based on invalid node data in response to the valid node data of the ith node device being less than or equal to the tth mean; and determining that the valid node data of the ith node device is the extremum in response to the valid node data of the ith node device being greater than or equal to an (n−1)th mean after (n−1)th round federated computation.