Motion-Based Facial Authentication Against Spoofed Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing facial recognition methods using 2D or 3D biometrics are vulnerable to fake images and require specialized devices and long measurement times, and biometric features like fingerprints can be faked or unavailable, leading to security challenges.
Innovation Solution
An authentication device that tracks facial muscle movements during recognition, utilizing neural networks to analyze motion features and gait, and incorporates a simulator to generate training data for improved precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If 2D or 3D biometric facial recognition is used, then authentication can be performed, but the system becomes vulnerable to fake images and 3D printed faces
Solution Approach 1:
The patent transitions from static 2D/3D facial recognition to dynamic facial muscle movement analysis. By capturing and analyzing the dynamic movements of facial muscles during expression changes, the system creates a living, breathing authentication that cannot be replicated by static fake images or 3D printed faces, thereby resolving the vulnerability to photorealistic forgeries.
Solution Approach 2:
The patent adds a temporal dimension to facial recognition by analyzing facial muscle movements over time. Instead of relying solely on the static geometric structure of the face (2D/3D), the system incorporates the dynamic behavior of facial muscles, creating a four-dimensional authentication that includes both spatial and temporal components, making it impossible for fake images to pass.
2Measurement precision
If traditional biometric devices are used for authentication, then identification can be performed, but specialized devices and long measurement times are required
Solution Approach 1:
The system uses the user's own facial muscle movements during natural expression changes as the authentication mechanism. Rather than requiring the user to deliberately perform specific actions or use specialized devices, the system captures authenticator information during normal social interactions, making the authentication process as fast and natural as having a conversation.
Solution Approach 2:
The authentication system works during any social interaction where the user's face is visible, eliminating the need for specialized authentication devices or separate measurement procedures. The same camera and processing system used for normal communication also performs authentication, making the process universally applicable and time-efficient.
3Reliability
If fingerprint or other biometric features are used, then authentication can be performed, but these features can be faked or are unavailable for some users
Solution Approach 1:
The system changes the authentication parameter from static physical features (fingerprints, facial geometry) to dynamic physiological behaviors (facial muscle movements). This parameter change makes authentication impossible to fake because real-time muscle movements during expression changes cannot be replicated by fingerprints or static 3D models, while also being available for all users regardless of fingerprint status.
Data Source
AI summary
Provided is an authentication device which includes an authentication moving picture obtaining unit which obtains authentication moving picture in which a person is captured for authentication and an authentication unit which authenticates the person based on the motions of the facial muscle of the person identified by analyzing the authentication moving picture. Provided is an authentication device which includes an authentication moving picture obtaining unit which obtains authentication moving picture in which a person is captured for authentication while walking and an authentication unit that authenticates the person based on the motions of the body muscles of the person identified by analyzing the authentication moving picture.