Targeted frequency estimation in local differential privacy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing data transmission methods lack effective mechanisms to reduce variance in frequency estimation for targeted frequency ranges, compromising data utility and security, particularly in local differential privacy protocols.

Innovation Solution

Adjust mechanism parameters to target specific frequency ranges, using hash functions and noise injection to generate encrypted output vectors, and employ multiple hash functions to minimize variance and enhance data utility.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If local differential privacy operations are applied to data before transmission, then data security and communication cost are improved, but data utility deteriorates due to increased variance in frequency estimation

Engineering Contradiction:
Improvedata securityVSAvoiddata utility
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The patent applies parameter changes by adjusting the inclusion probability parameter in the local differential privacy mechanism. By optimizing this parameter to minimize variance in frequency estimation for targeted ranges, the system achieves a balance between data security (through privacy mechanism) and data utility (through reduced variance). This directly resolves the contradiction by tuning the privacy mechanism parameters to preserve as much information utility as possible while maintaining security guarantees.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If noise is introduced into output vectors during local differential privacy processing, then data privacy is improved, but frequency estimation accuracy deteriorates

Engineering Contradiction:
Improvedata privacyVSAvoidfrequency estimation accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by differentiating the privacy mechanism based on frequency ranges. Instead of applying uniform noise addition to all data, the system introduces noise selectively and with different intensities for different frequency ranges. This allows high-frequency items to maintain better estimation accuracy while low-frequency items receive appropriate privacy protection, thereby resolving the contradiction between privacy and accuracy.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system changes the noise injection parameters dynamically based on the frequency range being analyzed. By adjusting the inclusion probability and noise magnitude according to the target frequency range, the system optimizes the trade-off between privacy protection and estimation accuracy for different parts of the data distribution.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If multiple hash functions are used in local differential privacy protocol, then frequency estimation variance is reduced, but system complexity increases

Engineering Contradiction:
Improvefrequency estimation varianceVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the frequency estimation problem into multiple segments corresponding to different frequency ranges. Each segment is processed with appropriate hash functions and privacy mechanisms tailored to its specific characteristics. This segmentation allows the system to reduce variance in each segment while managing overall complexity through structured processing.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4701124A1Targeted frequency estimation in local differential privacy
Publication Date: 2026.02.25 LEMON INC(GB)
  • EP4701124A1 patent drawingFigure 1
  • EP4701124A1 patent drawingFigure 2A~2B
  • EP4701124A1 patent drawingFigure 2C

AI summary

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reducing variance for target frequency ranges. One of the methods includes sending a request for data to each of a plurality of user devices; determining a target frequency range for the requested data; computing a value for an inclusion probability according to the target frequency range; providing the value for the inclusion probability to each of the plurality of user devices; receiving privatized messages from each of the plurality of user devices, each privatized message being generated according to the provided value for the inclusion probability; and analyzing the privatized data extracted from the received messages.