Private Set Union Data Protection via Vector Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The private set union (PSU) protocol faces challenges in ensuring data safety during transmission, as it may reveal additional information about the intersection between datasets, and the use of shuffling increases communication overhead, which is not efficiently managed by existing technologies.
Innovation Solution
A method is introduced that generates a Boolean vector for determining group similarities between datasets, followed by shuffling using a permutation function and masking vector, which reduces communication overhead while maintaining data safety by compressing the input vector length and preventing information leakage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If shuffling is used to prevent information leakage in PSU protocol, then data safety is improved, but communication overhead increases rapidly
Solution Approach 1:
The patent segments the large input vector into multiple smaller blocks, processing them in parallel. This reduces the communication overhead of shuffling by breaking down the monolithic data transmission into manageable chunks, thereby maintaining data safety while significantly reducing the total communication volume required.
Solution Approach 2:
The patent performs preprocessing on the input vector before the main PSU operation, organizing data into structured blocks with metadata. This preliminary action enables more efficient subsequent processing and reduces the communication overhead during the actual shuffle operation, as data is already optimally organized for parallel processing.
2Reliability
If hash table preprocessing is applied to PSU protocol, then data safety is improved, but additional information on intersection may be inferred
Solution Approach 1:
The patent introduces an intermediary masking mechanism that obscures the true values in the input vector during hash table preprocessing. By adding random masks and performing operations on masked data, the system maintains data safety through structured processing while preventing inference of additional intersection information that would otherwise be revealed.
3Ease of operation
If vector length is not compressed during shuffling, then processing simplicity is maintained, but communication costs increase
Solution Approach 1:
The patent divides the long input vector into multiple smaller blocks, each processed independently through the shuffling operation. This segmentation reduces communication costs by transmitting smaller data units in parallel, while maintaining processing simplicity through standardized block handling procedures that can be applied uniformly across all blocks.
Solution Approach 2:
The patent transforms the single-dimension problem of shuffling a long vector into a multi-dimensional problem by organizing data into blocks with spatial structure. This dimensional change enables parallel processing across multiple dimensions, reducing communication costs through efficient block-wise operations while maintaining operational simplicity through structured data organization.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure according to at least one embodiment provides a method for protecting data based on a private set union (PSU) protocol, performed by a first computing device for transmitting and receiving data to and from a second computing device, the method comprising: generating a Boolean vector for a first information set, based on whether or not each of a plurality of first groups that are groups of elements of the first information set stored in the first computing device is the same as each of a plurality of second groups that are groups of elements of a second information set stored in the second computing device, obtaining a random vector for the first information set as a result of shuffling performed by using the Boolean vector generated by the first computing device as an input and using order information and encryption information selected by the second computing device as inputs, and generating information on a union of the first information set and the second information set based on the random vector.