Image Injection Detection via Hue-Saturation Correlation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image verification methods in consumer devices are vulnerable to 'injection attacks' where spoofed or computer-generated images can be indistinguishable from real images, especially in mobile devices with limited computation resources, and often require computationally expensive pixel velocity tracking or spatial information correlation.

Innovation Solution

A computer-implemented method involving image frames conversion to hue-saturation color space, generating color distribution histograms, calculating correlations between test and control frames using zero-normalised cross-correlation, Chi-squared statistic, intersection correlation, and Bhattacharyya distance, and generating a verification output based on these correlations to classify images as matched or spoofed.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If pixel velocity tracking or spatial information correlation methods are used to detect injection attacks, then detection accuracy is improved, but computational cost increases making real-time execution impossible on mobile devices

Engineering Contradiction:

Improvedetection accuracyVSAvoidcomputational cost

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent extracts only the necessary color distribution information from image frames by converting to HSV color space and analyzing hue and saturation histograms, while discarding value/brightness information and spatial relationships. This extraction of essential features enables detection accuracy while dramatically reducing computational requirements for mobile device execution

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the parameter space by transforming images from RGB to HSV color space and then to 1D histograms, fundamentally altering how image data is represented. This parameter transformation reduces the dimensionality and complexity of the data while preserving the color distribution characteristics needed for injection attack detection

Inventive Principle:

Principle #35Parameter changes

2Reliability

If spatial information within image frames is correlated for verification, then verification reliability is improved, but security and privacy are compromised as reconstructive image data becomes accessible through the backend

Engineering Contradiction:

Improveverification reliabilityVSAvoidsecurity and privacy risk

Core Design Contradiction:

ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent removes spatial information from the analysis by converting 2D image frames into 1D color distribution histograms. This extraction eliminates the harmful aspect (spatial data accessibility) while retaining the useful aspect (color distribution verification) needed for reliable injection attack detection

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transitions from 2D spatial image data to 1D color distribution histograms, changing the dimensionality of the data representation. This dimensional transformation preserves verification capability through color distribution comparison while eliminating spatial information that could compromise security and privacy

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

3Measurement precision

If comprehensive image analysis methods are used to verify image authenticity, then verification accuracy is improved, but processing time increases preventing real-time verification

Engineering Contradiction:

Improveverification accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts only the essential color distribution characteristics from image frames by converting to HSV space and creating 1D histograms, discarding unnecessary spatial and brightness information. This selective extraction maintains verification accuracy while enabling real-time processing on mobile devices

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms the image representation parameters from full RGB pixel data to simplified HSV color distribution histograms, fundamentally changing how verification is performed. This parameter change reduces computational complexity while preserving the color-based verification accuracy needed for injection attack detection

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11232314B2Computer vision based approach to image injection detection

Publication Date: 2022.01.25 AZOULAY ROY

AI summary

A computer implemented method for image verification comprises receiving, by a processor, a plurality of image frames, comprising at least one test frame and one or more control frames; mapping, by the processor, each image frame onto a hue-saturation colour space; generating, based on the mapped image frames, a colour distribution for each image frame; calculating, by the processor, a correlation between the colour distributions generated for the at least one test frame and the one or more control frames; and generating, by the processor for output to a user, a verification output based on the calculated correlation.