Face Authentication Spoofing Detection via Head-Body Motion Consistency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face authentication systems are prone to spoofing, where a person can falsely pass authentication by displaying a printed or displayed image of another person, and high-resolution images can lead to inaccurate discriminator results, causing malicious individuals to bypass security measures.
Innovation Solution
An image processing apparatus that detects motion features of a person's head and body regions from time-series images and calculates an index value indicating the consistency between these features, using deep learning neural networks to determine the authenticity of the face image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If blink detection is used to identify spoofing, then spoofing can be detected in some cases, but high-resolution face images displayed on displays can be erroneously identified as authentic
Solution Approach 1:
The patent segments the motion analysis into two independent parts: head portion motion and body portion motion. By detecting and analyzing these separate motion features independently, the system can compare their consistency to identify spoofing, thereby resolving the contradiction between detecting spoofing and avoiding false positives with high-resolution images.
Solution Approach 2:
The patent transitions from two-dimensional face image analysis to three-dimensional motion analysis by incorporating temporal dimension (time-series images) and spatial dimension (both head and body portions). This dimensional expansion enables the system to detect spoofing through motion inconsistency that cannot be detected in static or single-region analysis.
2Measurement precision
If discriminator accuracy is improved for high-resolution images, then authentication precision increases, but spoofing detection capability decreases
Solution Approach 1:
The patent divides the authentication process into two independent analysis streams: one for face image discrimination and another for motion consistency verification. This segmentation allows each stream to optimize for its specific function without compromising the other, resolving the contradiction between discrimination accuracy and spoofing detection reliability.
Solution Approach 2:
The patent introduces motion consistency analysis as an intermediary verification layer between image capture and final authentication decision. This intermediary mechanism checks the consistency between head and body motions, providing an additional safeguard that maintains authentication precision while enhancing spoofing detection reliability.
3Ease of operation
If only head portion motion is analyzed for spoofing detection, then detection simplicity is maintained, but spoofing identification accuracy decreases
Solution Approach 1:
The patent segments motion detection into head portion and body portion analysis, where each segment is processed independently through similar detection pipelines. This segmentation maintains the simplicity of individual detection processes while improving overall accuracy through the combination and consistency check of multiple segments.
Data Source
AI summary
An image processing device is provided with: a movement detection unit—for detecting, from time-series images, a first feature relating to the first movement of a head portion of a person and a second feature relating to the second movement of a body portion, which is a part of the person other than the head portion; and an index value calculation unit for calculating an index value which indicates the degree of consistency between the first feature relating to the first movement of the head portion of the person and the second feature relating to the second movement of the body portion of the person.