Renyi Divergence Privacy Budget Tracking for Database Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing database systems face performance bottlenecks due to high computing resource requirements for determining accurate aggregate privacy spends in differentially private systems, which limits the frequency of database queries and compromises analytical utility.
Innovation Solution
A differentially private security system that determines a noise type for database queries based on probabilistic privacy loss, using a Renyi differential privacy curve, and adjusts privacy parameters to ensure queries do not exceed a privacy budget, thereby blocking or executing queries accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If accurate aggregate privacy spend determination is performed using traditional differential privacy methods, then privacy protection is ensured, but computing resource requirements increase significantly
Solution Approach 1:
The patent uses pre-computed, disposable privacy spend bounds for each query type that can be quickly looked up rather than computed in real-time. These bounds are calculated once and stored, then reused for similar queries, eliminating the need for expensive repeated computations while maintaining privacy guarantees.
Solution Approach 2:
The system performs preliminary computation of privacy spend bounds before queries are executed. By pre-calculating and storing privacy spend information for different query types and data subsets, the system avoids performing these expensive computations during actual query processing, thus reducing real-time computing resource requirements.
2Reliability
If traditional differential privacy query processing is used, then privacy guarantees are maintained, but database query frequency is limited due to performance bottlenecks
Solution Approach 1:
The patent segments the privacy budget management into independent, query-type-specific components. Each query type has its own pre-computed privacy spend bound, allowing parallel processing and independent optimization of different query types without interfering with each other, thus enabling higher query throughput.
Solution Approach 2:
The system creates and stores copyable privacy spend bound representations for different query types. These pre-computed bounds can be quickly copied and applied to multiple similar queries without re-computation, significantly increasing the frequency at which queries can be processed while maintaining privacy guarantees.
3Reliability
If privacy spend bounding is performed for each query, then privacy budget control is achieved, but computing overhead increases and analytical utility is compromised
Solution Approach 1:
The patent changes the parameter representation from detailed per-query privacy spend calculations to aggregated privacy spend bounds based on query type and data subset characteristics. This parameter transformation enables efficient privacy budget control through simple comparisons rather than complex computations, preserving analytical utility.
Solution Approach 2:
The pre-computed privacy spend bounds serve multiple functions: they control privacy budget expenditure, enable quick query feasibility determination, and support various query types through a unified framework. This multi-functionality eliminates the need for separate privacy management mechanisms for different query types, reducing overall computing overhead.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A differentially private security system communicatively coupled to a database storing restricted data receives a database query from a client. The database query includes a relation indicative of data to perform the query upon and at least one privacy parameter indicative of a level of differential privacy with which to perform the query. The differentially private security system determines a noise type for the query. The differentially private security system determines a representation of probabilistic privacy loss for the query based on the determined noise type. The differentially private security system determines a privacy spend for the query using the generated representation of probabilistic privacy loss. The differentially private security system determines whether the determined privacy spend exceeds a privacy budget associated with the client.