Facial Recognition Anti-Spoofing via Dynamic Projection Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing live recognition-based authentication systems are vulnerable to video or image attacks, where illegitimate users simulate the actions of legitimate users, leading to security breaches.
Innovation Solution
The system employs a method where projection information, such as changing brightness or color, is displayed on a target object, and images are captured before and after the change, with image difference data analyzed to determine if the target is a virtual or living object, thereby enhancing authentication security.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If facial recognition system uses conventional image capture and recognition techniques, then the system is simple and easy to operate, but the system becomes vulnerable to photo and video attacks
Solution Approach 1:
The system performs preliminary actions by capturing a first image before displaying projection information and a second image after displaying projection information. This preliminary capture of images at different states enables subsequent comparison to detect whether the target is a real person or a virtual object, thereby improving authentication security before the actual authentication decision is made.
Solution Approach 2:
The system transitions from conventional single-state facial recognition to multi-state recognition by capturing images before and after projection information display. This adds a temporal and visual dimension to the recognition process, allowing the system to detect subtle changes in light reflection that differentiate real faces from virtual representations.
2Reliability
If the system requires user actions like nodding or eye-blinking for authentication, then the system can distinguish real persons from static images, but the system becomes vulnerable to video playback and software simulation attacks
Solution Approach 1:
The system changes visual parameters by displaying projection information with different brightness, color, or graphical characteristics on the screen. By capturing images before and after these parameter changes and comparing them, the system can detect whether the target object (real face or virtual representation) responds appropriately to visual stimuli, thereby distinguishing real persons from video or image attacks.
Solution Approach 2:
The system implements feedback by displaying projection information on a screen that the target object can see, then capturing how the target object's appearance changes in response. This feedback loop allows the system to verify that the target is a real person who naturally responds to visual stimuli rather than a pre-recorded video or static image.
3Measurement precision
If the system uses only single image capture for recognition, then the processing speed is fast and simple, but the measurement precision of determining whether the target is real or virtual is insufficient
Solution Approach 1:
The system performs preliminary image captures at two distinct time points (before and after projection information display) to gather sufficient data for accurate real-vs-virtual detection. This preliminary data collection enables precise measurement without requiring multiple iterative attempts, thereby maintaining authentication speed while improving detection precision.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Facial recognition-based authentication comprises obtaining a first image of a target object, updating projection information associated with a display by a display device, obtaining a second image of the target object, the second image being an image of the target object after the projection information is updated, obtaining an image difference data based at least in part on the first image and the second image, and determining whether the target object is a virtual object based at least in part on the image difference data.