Face Spoofing Detection via Siamese Neural Network Branches
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated systems for verifying ID photographs are limited in detecting real faces from fraudulent representations, particularly in preventing 'face spoofing' and ensuring compliance with varying national regulations, and are complex and not suitable for devices with limited computing capacity.
Innovation Solution
A method using a computing device to acquire two images of a person's face within a short time frame, extract facial cues, propagate them through Siamese branches of a neural network, and evaluate the random or non-random nature of face movement to determine if the image is suitable for use as an ID photograph, which can be implemented on devices with limited computing capacity like smartphones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If complex fraud detection methods are used to detect face spoofing, then detection reliability is improved, but device complexity increases
Solution Approach 1:
The method segments the face detection task into multiple independent components: facial cue extraction, Siamese branch processing, cost function evaluation, and classification. This segmentation allows each component to be optimized independently and reduces overall system complexity while maintaining high detection reliability through specialized processing at each stage.
Solution Approach 2:
The patent introduces intermediary representations (facial cue vectors, Siamese branch outputs) that mediate between the input images and the final fraud detection decision. These intermediaries simplify the processing by breaking down the complex relationship between images and fraud detection into manageable transformation steps, reducing computational complexity while preserving detection accuracy.
2Productivity
If automated verification is implemented to ensure compliance, then productivity is improved, but reliability deteriorates due to increased fraud risk
Solution Approach 1:
The system implements feedback mechanisms where the classification output is used to verify compliance and detect fraud attempts. The cost function evaluation provides feedback on the likelihood of fraud, allowing the system to adjust its verification decisions in real-time, maintaining both high productivity through automation and high reliability through continuous fraud detection feedback.
Solution Approach 2:
The method performs preliminary actions by extracting facial cues and evaluating them through Siamese branches before making the final fraud detection decision. This preliminary processing allows the system to identify potential fraud attempts early in the verification process, maintaining high productivity while improving reliability by catching fraud attempts before they are approved.
3Ease of operation
If simple processing methods are used on limited computing devices, then ease of operation is improved, but measurement precision deteriorates
Solution Approach 1:
The patent changes parameters such as the number of facial cues extracted, the depth of Siamese branch processing, and the complexity of the cost function evaluation to optimize the balance between computational resources and detection precision. By adjusting these parameters, the system achieves high fraud detection precision on devices with limited computing capacity, making ID photograph verification accessible to a wider range of devices while maintaining accurate fraud detection.
Data Source
AI summary
A method for determining whether an image of a person's face is suitable for use as an ID photograph comprises a step of acquiring a first image and a second image of the person's face. The method also comprises a step of propagating facial cues from the first image and facial cues from the second image into two Siamese branches of a main neural network. A computer program may be employed to implement the method, and a system may be configured to implement the method.