Frequency-Band Image Authenticity Detection for Fraud Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing non-face-to-face authentication methods for financial services, such as those using captured images of ID cards, are vulnerable to fraud due to the ease with which criminals can steal and manipulate these images, leading to increased financial crimes.
Innovation Solution
An image determination method and system that converts images into a frequency domain, extracts data using multiple band masks with overlapping regions, and performs inverse transformation to generate and analyze features from both global and local image regions, utilizing a deep learning model for authenticity assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If non-face-to-face authentication using captured images is used, then convenience and accessibility of financial services are improved, but security and vulnerability to fraud increase
Solution Approach 1:
The patent transforms images from the spatial domain to the frequency domain using Fourier transform, changing the representation parameters of image data to reveal hidden characteristics and patterns that are not visible in the original spatial domain, thereby enabling more accurate authenticity determination
Solution Approach 2:
The patent divides the frequency-domain image into multiple frequency bands using band masks, extracting and analyzing specific frequency components separately to identify characteristics specific to different types of images (original, printed, displayed), thereby improving fraud detection accuracy
2Speed
If simple image authentication is used, then processing speed is improved, but accuracy of authenticity determination deteriorates
Solution Approach 1:
The patent segments the frequency spectrum into multiple bands and extracts features from each band separately, then combines these features for comprehensive authenticity determination, achieving high accuracy without excessive processing time by focusing on characteristic frequency regions
Solution Approach 2:
The patent applies different band masks to extract specific frequency characteristics from different regions of the frequency spectrum, analyzing local frequency patterns that are indicative of image authenticity, thereby improving measurement precision through targeted analysis
3Measurement precision
If frequency domain analysis with multiple band masks is applied, then authenticity determination accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent divides the complex frequency-domain analysis into multiple manageable steps by applying different band masks to extract specific frequency characteristics, making the complex authentication process more systematic and computationally efficient through structured analysis
Solution Approach 2:
The patent uses frequency-domain representation as an intermediary between the original spatial-domain image and the final authenticity determination, transforming the image into frequency components that are easier to analyze and compare for fraud detection
Data Source
AI summary
There is provided an image determination method and system. The image determination method includes: converting a target image in a spatial domain into an image in a frequency domain; extracting first image data associated with a first frequency band by applying a first band mask to the frequency-domain image; extracting second image data associated with a second frequency band by applying a second band mask to the frequency-domain image, wherein the first and second frequency bands have an overlapping band region; generating a plurality of images by inverse-transforming the first image data and the second image data into the spatial domain; and determining authenticity of the target image based on features extracted from the plurality of images. According to the image determination method and system, the authenticity of the target image can be accurately determined.