Image-Based Data Encoding for Confidential Consistency Checks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing data analysis methods struggle to efficiently compare and maintain confidentiality of sensitive data sets while determining their consistency, as direct data sharing risks privacy breaches.

Innovation Solution

Encoding data values into images and using an image-classification machine learning model to generate numerical output values, allowing secure comparison without revealing the underlying data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If data values are directly shared and compared between computing devices, then data consistency can be evaluated efficiently, but confidentiality and privacy of sensitive data are compromised

Engineering Contradiction:
Improvedata consistency evaluationVSAvoidprivacy breach
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent introduces an intermediary encoding process that transforms sensitive data values into image representations before transmission. The image-classification machine learning model acts as a mediator that processes these encoded images and outputs consistency evaluation results without exposing the underlying sensitive data values, thus resolving the contradiction between reliable consistency evaluation and privacy protection

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent creates a copy of the data in the form of images that represent the data values. These image copies are transmitted and processed instead of the original sensitive data, allowing consistency evaluation while the original confidential information remains protected on the source devices

Inventive Principle:
Principle #26Copying

2Object-affected harmful factors

If data is encoded into images and processed through machine learning models, then confidentiality is preserved, but the complexity of the system increases

Engineering Contradiction:
Improveconfidentiality preservationVSAvoidsystem complexity
Core Design Contradiction:
Object-affected harmful factorsVSDevice complexity

Solution Approach 1:

The patent employs image-classification machine learning models that are universally applicable for processing encoded data images. These models serve multiple functions: they process the encoded image data, evaluate consistency between different data sets, and maintain confidentiality simultaneously, thereby managing system complexity through multi-functional components

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Productivity

If traditional data comparison methods are used, then the process is simple and fast, but sensitive data must be exposed for comparison

Engineering Contradiction:
Improvedata comparison efficiencyVSAvoiddata privacy loss
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The encoding and image-processing pipeline serves as an intermediary layer that enables efficient comparison operations to be performed on encoded representations rather than raw sensitive data. The machine learning model processes these encoded images rapidly while the underlying data remains confidential, thus maintaining productivity without sacrificing privacy

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4657860A1Data encoding using images and machine learning models
Publication Date: 2025.12.03 SAP SE
  • EP4657860A1 patent drawingFigure 1
  • EP4657860A1 patent drawingFigure 2
  • EP4657860A1 patent drawingFigure 3

AI summary

Summarizing the invention, a computer-implemented method is provided. The computer-implemented method comprises: obtaining, by a first computing device, a plurality of first data values; creating, by the first computing device, a first image comprising a plurality of first pixels, wherein the number of first pixels is equal to the number of first data values and wherein each first data value of the plurality of first data values is assigned to a respective first pixel of the plurality of first pixels; providing the first image as input to an image-classification machine learning model to obtain a first numerical output value; obtaining, by a second computing device, a plurality of second data values; creating, by the second computing device, a second image comprising a plurality of second pixels, wherein the number of second pixels is equal to the number of second data values and wherein each second data value of the plurality of second data values is assigned to a respective second pixel of the plurality of second pixels; providing the second image as input to the image-classification machine learning model to obtain a second numerical output value; evaluating the first numerical output value and the second numerical output value to determine whether the plurality of first data values is consistent with the plurality of second data values.