Anonymous Multi-Conversion PSI with Differential Privacy Padding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing secure multi-party computation and communication protocols fail to protect membership privacy during private set intersection operations, particularly in scenarios where a single user performs multiple valuable actions, leading to potential membership leakage and exposure.
Innovation Solution
Implementing a differential privacy protocol with padding elements and noise addition to ensure that the intersection size is random and private, preventing attackers from determining user membership, while maintaining user anonymity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional PSI protocols are used to compute intersection of datasets, then the intersection size can be accurately computed, but user membership privacy is leaked and exposed
Solution Approach 1:
The patent introduces padding elements as an intermediary component that mediates between the need for accurate intersection computation and privacy protection. These padding elements are added to datasets before PSI operations, acting as a buffer that prevents direct exposure of real user memberships while still allowing the intersection protocol to function. The padding elements absorb the privacy risk by being indistinguishable from real elements during the intersection computation.
Solution Approach 2:
The patent changes the parameter of dataset composition by adding padding elements with specific properties (different distribution characteristics) to the original datasets. This parameter change transforms the datasets from pure user membership lists to mixed datasets containing both real and synthetic elements, thereby altering the privacy properties while maintaining the functional capability for intersection computation.
2Object-affected harmful factors
If padding elements are added to datasets to protect privacy, then user membership privacy is protected, but the complexity of the PSI protocol increases
Solution Approach 1:
The patent applies preliminary action by adding padding elements to datasets before the PSI protocol execution. This preprocessing step prepares the datasets in advance with the necessary privacy protections, so that during the actual PSI computation, the protocol can proceed with standard operations without needing complex modifications. The complexity is front-loaded into a simple padding addition step rather than being distributed throughout the entire protocol.
3Object-affected harmful factors
If noise is added to make intersection size random and private, then differential privacy is achieved, but the precision of intersection measurements is reduced
Solution Approach 1:
The patent creates copies of dataset elements in the form of padding elements that replicate the structure and appearance of real elements but contain no actual membership information. These synthetic copies serve as placeholders that allow the intersection protocol to produce results that appear precise but are actually randomized through the differential privacy mechanism. The copying approach enables privacy protection while maintaining the functional form of intersection computation.
Data Source
AI summary
Systems and techniques described herein provide private set intersection (PSI) algorithms or protocols that improve the identification of “multi-conversion” within usage datasets while also keeping the users in the datasets anonymous during PSI operations. In some implementations, a first and a second dataset are dispatched. A first intersection operation is performed based on the first dataset and the second dataset. A second intersection operation is then performed based on the result of the first intersection operation. A third dataset is generated based on the first and second intersection operations, where the third dataset includes one or more identifications reflecting a multi-conversion event.