Private Set Intersection with Bias Parameters for DP Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for differentially private set intersection (PSI) in client-server models face challenges in balancing data privacy protection with precision and recall, as introducing noise for privacy can degrade these metrics, and malicious users can exploit shared data to derive sensitive information.
Innovation Solution
A method involving a server that determines bias parameters based on ε-differential privacy (ε-DP) to protect data privacy by iteratively updating the intersection dataset using random number generators with specific probabilities, ensuring optimal precision and recall while maintaining privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If differential privacy protection is applied to the intersection dataset by introducing noise, then data privacy is protected, but data precision and recall are degraded
Solution Approach 1:
The patent changes the parameters of noise introduction by using bias parameters (p00, p01, p10, p11) that are derived from the differential privacy parameter ε. These bias parameters control the probability of adding or removing data entries, allowing optimization of the balance between privacy protection and data precision/recall. The specific parameter transformation from ε to bias parameters enables fine-grained control over the privacy-utility tradeoff.
Solution Approach 2:
The patent introduces dynamic adjustment mechanisms where the bias parameters can be adapted based on the specific characteristics of the dataset and the desired level of privacy protection. The system dynamically determines which data entries to add or remove based on randomized responses with probabilities controlled by the bias parameters, creating a flexible and adaptive privacy protection mechanism rather than a static noise addition approach.
2Reliability
If existing differential privacy methods are used for set intersection, then privacy protection is provided, but computing resources are excessively consumed due to multiple iterations
Solution Approach 1:
The patent performs preliminary computation by determining all bias parameters (p00, p01, p10, p11) before processing the intersection dataset. This preliminary determination of parameters based on the differential privacy requirement allows the system to execute a single-pass iteration through the data, eliminating the need for multiple iterative passes that consume excessive computing resources. The preliminary parameter setup enables efficient one-shot computation of the differentially private set intersection.
3Measurement precision
If the intersection dataset is returned without differential privacy protection, then data precision and recall are maintained, but malicious users can perform adaptive attacks to derive sensitive information
Solution Approach 1:
The patent introduces randomized responses with bias parameters as an intermediary mechanism between the original intersection dataset and the returned result. This intermediary layer of probabilistic addition and removal of data entries acts as a shield that prevents malicious users from performing adaptive attacks, while still allowing legitimate data analysis to proceed with acceptable precision and recall. The intermediary mechanism transforms the raw intersection data into a differentially private version that resists attacks.
Data Source
AI summary
Methods, systems, and apparatuses, including computer programs encoded on computer storage media, for data communication with differentially private intersection dataset. On example method includes: determining a plurality of bias parameters based on a parameter for establishing a differential privacy (DP) data protection; determining an intersection dataset based on client data from a client device and server data from the server device; updating the intersection dataset based on the client data and the plurality of bias parameters; and returning the updated intersection data having the DP data protection to the client device.