Database Privacy Mechanism Using Randomized Response and Exponential Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating outputs from databases while ensuring differential privacy are inefficient, requiring large samples and many trials due to high standard errors from randomized response algorithms and complexity of quality function-based mechanisms like k-anonymity.
Innovation Solution
A method that preprocesses database elements using a randomized response algorithm and then applies a differential privacy algorithm based on a quality function, specifically an exponential mechanism, to reduce computing resources and improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a randomized response algorithm is used to provide differential privacy, then client-based privacy protection is achieved without requiring a trusted third-party server, but larger standard errors occur leading to reduced power and requiring comparatively large samples to generate a dataset with sufficient accuracy
Solution Approach 1:
The patent segments the differential privacy mechanism into two distinct parts: (1) a randomized response algorithm for initial privacy protection, and (2) a quality function-based mechanism for selecting high-utility elements. This segmentation allows each component to specialize - the randomized response provides foundational privacy while the quality function optimizes accuracy, resolving the contradiction between privacy reliability and measurement precision
Solution Approach 2:
The patent changes the parameters of the differential privacy mechanism by introducing a quality function that evaluates elements based on multiple criteria including accuracy metrics. By adjusting the quality function parameters and using an exponential mechanism with carefully selected privacy budget allocation, the system achieves both strong privacy protection and high output accuracy, overcoming the limitation of traditional randomized response methods
2Measurement precision
If a differential privacy mechanism based on a quality function is applied directly to database elements, then the best element with highest utility can be selected while preserving differential privacy, but significantly more computing resources and trials are required
Solution Approach 1:
The patent applies preliminary action by first processing the database through a randomized response algorithm to create a pre-processed dataset with embedded privacy protection. This preliminary step reduces the complexity of subsequent quality function evaluation, as the privacy guarantees are already established, allowing the system to find high-utility elements with fewer trials and less computational effort
Solution Approach 2:
The patent introduces an intermediary mechanism - the exponential mechanism with a carefully designed quality function - that mediates between the randomized response output and the final selection. This intermediary efficiently evaluates elements based on multiple criteria and selects the optimal element while maintaining differential privacy, significantly improving computing efficiency compared to brute-force approaches
3Reliability
If private data publishing mechanisms like k-anonymity are used to achieve differential privacy, then privacy protection is provided, but the mechanisms become comparatively complex requiring multiple trials to achieve the desired privacy level
Solution Approach 1:
The patent extracts the essential privacy protection function from complex mechanisms like k-anonymity and implements it through a simpler randomized response algorithm combined with a quality function-based selection. By taking out only the necessary privacy-guaranteeing elements and discarding the complex iterative procedures of traditional mechanisms, the system achieves equivalent or superior privacy protection with reduced complexity and fewer trials
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a method for generating an output in response to a query on a database containing data pertaining to a plurality of privacy principles, wherein the method (10) comprises the steps of: - generating at least one data quantity based on the data contained in the database using a randomized response algorithm (2,11); and - generating the output based on the at least one data quantity using a differential privacy algorithm that is based on a quality function (3,12).