Single Feature Vector Encryption for Image Privacy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing feature extraction methods produce multiple feature vectors for images, which are not amenable to secure encryption, making them susceptible to attacks and compromising image privacy during comparison.

Innovation Solution

Generating a single feature vector from multiple feature vectors extracted from an image, which is then encrypted, resulting in higher entropy and improved security for comparison purposes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If multiple feature vectors are extracted from an image, then feature representation completeness is improved, but security against attacks deteriorates

Engineering Contradiction:
Improvefeature representation completenessVSAvoidsecurity against attacks
Core Design Contradiction:
Loss of informationVSReliability

Solution Approach 1:

The patent combines multiple feature vectors extracted from an image into a single aggregated feature vector. This merging process maintains the comprehensive feature representation while creating a unified structure that can be securely encrypted and compared, thereby resolving the contradiction between feature completeness and security.

Inventive Principle:
Principle #5Merging (Combining)

2Reliability

If feature vectors are encrypted for security, then image privacy is protected, but comparison capability deteriorates

Engineering Contradiction:
Improveimage privacy protectionVSAvoidcomparison capability
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The patent applies encryption transformation to the aggregated feature vector, changing its parameter representation while preserving the essential comparison capability. The encryption scheme allows for secure comparison of encrypted feature vectors without requiring decryption, thus maintaining both privacy protection and comparison functionality.

Inventive Principle:
Principle #35Parameter changes

3Reliability

If multiple feature vectors are individually encrypted, then security entropy is increased, but operational complexity deteriorates

Engineering Contradiction:
Improvesecurity entropyVSAvoidoperational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent aggregates multiple feature vectors into a single feature vector before encryption, which simplifies the encryption operation to a single process rather than multiple individual encryptions. This approach maintains security through the entropy of the aggregated data while reducing operational complexity by eliminating the need to manage multiple encryption operations.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3410352B1Feature vector generation and encryption
Publication Date: 2023.08.23 FUJITSU LTD
  • EP3410352B1 patent drawingFigure 1A
  • EP3410352B1 patent drawingFigure 1B
  • EP3410352B1 patent drawingFigure 1C

AI summary

A method may include extracting feature-data from an image which includes location data that indicates multiple locations within the image and multiple feature vectors that each correspond to one of the multiple locations. Each feature vector may represent a feature of an object in the image. The method may include determining a feature-data plot based on relative positions of the multiple locations and selecting a grid for the feature-data plot. The method may include generating a first representative feature vector for a first cell of the grid based on a first feature vector set and generating a second representative feature vector for a second cell of the grid based on a second feature vector set. The method may include generating a single feature vector that represents the object in the image based on the first representative feature vector and the second representative feature vector and encrypting the single feature vector.