3D Face Reconstruction for Biometric Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing biometric identity verification systems struggle with accurately distinguishing between genuine human faces and spoofed representations, often requiring complex equipment and user movements, and are prone to noise and limitations in 3D reconstruction, especially with common feature extractors like SIFT, leading to suboptimal fraud detection.
Innovation Solution
A system using a mobile device with a 3D facial reconstruction and deep learning framework that combines facial attributes and image quality characteristics to generate embeddings for fraud detection, allowing for robust fraud assessment with minimal user interaction and common consumer devices, utilizing a three-dimensional reconstruction from video captured at multiple angles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If common feature extractors like SIFT are used for 3D reconstruction, then the system can work with common consumer devices, but the reconstruction quality degrades due to significant noise
Solution Approach 1:
The patent changes the parameters of the feature extraction process by using deep learning-based feature extractors instead of traditional SIFT, and by modifying the 3D reconstruction algorithm to be more robust to noise from consumer devices. This resolves the contradiction by maintaining compatibility with common devices while improving reconstruction quality through parameter optimization.
Solution Approach 2:
The patent replaces traditional mechanical/optical 3D reconstruction methods with deep learning-based approaches. The neural network-based feature extractors and reconstruction algorithms substitute for conventional computer vision techniques, enabling high-quality 3D reconstruction from consumer device video without requiring specialized hardware.
2Measurement precision
If sophisticated 3D reconstruction algorithms are used to improve fraud detection accuracy, then detection precision improves, but the system becomes more complex and requires specialized equipment
Solution Approach 1:
The patent replaces complex traditional 3D reconstruction algorithms with deep learning-based approaches that are more efficient and easier to implement. The neural network automatically learns features and performs reconstruction in an integrated framework, reducing system complexity while improving fraud detection accuracy through better feature extraction and noise robustness.
Solution Approach 2:
The patent merges feature extraction, 3D reconstruction, and fraud detection into a unified deep learning framework. This integration eliminates the need for separate complex modules and specialized equipment, achieving high detection accuracy through a consolidated system that processes video data end-to-end.
3Reliability
If existing liveness detection techniques are used, then some fraud detection capability is provided, but they require complicated user actions and movement-intensive interactions
Solution Approach 1:
The patent enables the system to perform fraud detection using passive observation of natural user behavior during normal video capture. The deep learning model automatically analyzes facial movements, blood flow patterns, and other physiological signals without requiring the user to perform specific actions, making the process as easy as taking a selfie while maintaining high detection reliability.
Solution Approach 2:
The patent replaces active liveness detection methods requiring user actions with passive physiological signal analysis. Instead of asking users to perform movements, the system uses deep learning to detect subtle physiological indicators like blood flow and micro-movements that occur naturally, greatly simplifying user interaction while maintaining or improving detection accuracy.
Data Source
AI summary
Described herein are computerized methods and systems for detecting fraud during biometric identity verification. A mobile device captures video comprising a plurality of frames of a person's face. The mobile device extracts from the video frames comprising at least two frames of the person's face from different angles. The mobile device creates a reconstruction of the person's face using the extracted frames and derives signals associated with features of the person's face. The mobile device generates an embedding for each extracted frame using the extracted frames, the three-dimensional reconstruction, and the signals. The mobile device calculates, for each extracted frame, a fraud confidence value based upon the embedding for the extracted frame, attributes of the person's face, and image quality attributes of the extracted image. The mobile device computes a fraud detection decision for the extracted frames based upon the fraud confidence values and the embeddings.