Identity Image Fraud Detection Through Foreground Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting fraudulent image swaps in identity documents, such as portrait photo substitution, are inefficient as they require multiple frames and fail to detect subtle holograms or colorless features.
Innovation Solution
A fraudulent image detector using convolutional neural networks to analyze a single image, segmenting foreground and backdrop regions, and computing probabilities for security patterns, outputs a fraud detection score based on threshold comparisons.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If chromaticity statistics and hue variation analysis are used to detect fraud, then detection capability is provided, but the system requires multiple frames and fails to detect holograms in pale or colorless regions
Solution Approach 1:
The patent segments the image into foreground region (containing personal identifiable information) and backdrop region (containing security patterns). This segmentation allows independent analysis of each region's characteristics, enabling detection of fraud in a single frame without requiring multiple frames for comparison.
Solution Approach 2:
The patent analyzes local quality differences between foreground and backdrop regions, specifically examining chromaticity statistics and hue variation patterns in different spatial locations. This local analysis enables detection of subtle inconsistencies that indicate fraud, including holograms in pale or colorless regions, without needing multiple frames.
2Reliability
If chromaticity statistics and hue variation analysis are used to detect fraud, then detection capability is provided, but the system fails to detect holograms appearing near pale or colorless regions
Solution Approach 1:
The patent applies local quality analysis by examining chromaticity statistics and hue variation patterns specifically in different spatial regions. This allows detection of holograms in pale or colorless regions by analyzing the local color distribution characteristics rather than relying on global color analysis.
Solution Approach 2:
The patent introduces a spatial dimension to the analysis by segmenting the image into foreground and backdrop regions and analyzing each region's chromaticity statistics independently. This dimensional approach enables detection of holograms in pale or colorless regions by examining color distribution across different spatial locations.
3Reliability
If pixel-wise chromaticity statistics analysis is used, then fraud detection is enabled, but subtle evidence invisible to human verifiers is missed
Solution Approach 1:
The patent segments the image into foreground and backdrop regions, enabling analysis of subtle chromaticity variations in each region. This segmentation allows detection of subtle evidence invisible to human verifiers by examining local color statistics that would be missed in global analysis.
Solution Approach 2:
The patent replaces human visual verification with automated computer-based chromaticity statistics analysis. This substitution enables detection of subtle evidence that is invisible to human verifiers by using computational methods to analyze color distribution patterns at the pixel level.
Data Source
AI summary
A fraudulent image detector including a first segmenter configured to compute first probability data from an image showing personal identifiable information of an individual, the first probability data indicating, for each pixel of the image, a probability that the pixel shows a security pattern, a second segmenter configured to compute second probability data from the image, the second probability data indicating, for each pixel of the image, a probability that the pixel is part of a foreground region showing the personal identifiable information or part of a backdrop region showing no personal identifiable information, and a classifier configured to compute score data from the first probability data and the second probability data.