Face Liveness Scoring Fusion for Presentation Attack Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face recognition systems are vulnerable to various types of presentation attacks (PAs) and lack adaptability to new forms of spoofing, with passive liveness determination methods failing to detect a broad range of PAs and being highly dependent on video frame quality.
Innovation Solution
A system utilizing two scoring modules in parallel, each with its own processing, combined with a quality filtering module to reject low-quality frames, enhances liveness determination by using convolutional neural networks (CNNs) for spatial features and additional quality features like depth maps and skin texture, along with a fusion module to merge scores for accurate liveness assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If passive liveness determination methods are used to avoid obtrusiveness, then user experience is improved, but detection accuracy against various presentation attacks deteriorates
Solution Approach 1:
The system segments the liveness determination task into multiple independent scoring modules, each specialized for detecting specific types of presentation attacks. This allows the system to maintain passive operation while achieving comprehensive detection coverage through the combined output of multiple specialized detectors.
Solution Approach 2:
The system creates a universal liveness determination framework that can detect multiple types of presentation attacks (photo, video, mask, replay) using a single passive system. The multi-functional scoring modules collectively provide broad detection capability without requiring active user cooperation.
2Device complexity
If a single liveness determination model is used to simplify the system, then device complexity is reduced, but adaptability to different types of presentation attacks deteriorates
Solution Approach 1:
The system divides the liveness determination function into multiple specialized scoring modules, each trained to detect specific presentation attack types. This segmentation allows each module to be relatively simple while the collective system achieves high adaptability across diverse attack scenarios.
Solution Approach 2:
The system merges the outputs of multiple specialized scoring modules through a fusion mechanism to produce the final liveness determination. This combining approach maintains individual module simplicity while achieving comprehensive adaptability through their integrated decision-making.
3Measurement precision
If liveness determination is highly dependent on video frame quality to improve accuracy, then detection precision is improved, but reliability under varying capture conditions deteriorates
Solution Approach 1:
The system applies quality filtering at the local frame level, assessing and weighting individual frames based on their quality metrics. This allows the system to maintain high detection precision by focusing on quality frames while preserving reliability by appropriately handling or discarding low-quality frames rather than failing entirely.
Data Source
AI summary
A system for determining liveness of a target person comprising a frame capture module, a face detection module and a frame quality module configured to determine at least one quality feature from each frame. A quality filtering module is configured to reject or accept each frame based on a comparison between a predefined capture condition and a first quality feature. A first scoring module is arranged, and to determine a first score based on the detected face of a frame, if it is accepted. A second scoring module is arranged to determine a second score based on at least one second quality feature extracted from a frame, if it is accepted. A fusion module is configured for attributing a final score representative of liveness of the target person based on the first and second scores.