Biometric Spoof Detection Using Skip Connections and Attention Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing biometric systems struggle to accurately distinguish between genuine and fake biometric information, particularly when fake fingerprint patterns are used, leading to false authentication and increased computational complexity in detection.
Innovation Solution
A neural network architecture incorporating skip connections and attention blocks in convolution blocks to enhance spoofing detection performance while minimizing computational complexity, using a small-sized neural network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a typical fingerprint recognizing device is used to compare fingerprint images, then authentication can be performed, but the device cannot distinguish finely fabricated fake fingerprint patterns from genuine fingerprints, leading to false recognition
Solution Approach 1:
The patent segments the fingerprint analysis into multiple feature extraction channels within a convolutional neural network. Different channels extract different characteristics (e.g., ridge patterns, minutiae, texture features) that collectively enable discrimination between genuine and spoofed fingerprints, addressing the limitation of single-feature recognition systems
Solution Approach 2:
The patent employs a composite approach by integrating multiple convolution blocks with different kernel sizes and configurations to process fingerprint images. This multi-layered composite structure combines various feature extraction capabilities to create a robust spoof detection system that overcomes the weaknesses of individual recognition methods
2Reliability
If conventional fingerprint recognition methods are used, then processing is simple, but the system falsely accepts spoofed biometric information as genuine
Solution Approach 1:
The patent implements dynamic feature extraction through convolutional neural network blocks that adaptively learn and extract relevant fingerprint features. The network dynamically adjusts feature weights and combinations during training, enabling it to capture complex spoof detection patterns without requiring manually engineered complex rules
Solution Approach 2:
The patent introduces an intermediary convolutional neural network model that acts as a mediator between raw fingerprint images and authentication decisions. This intermediary processing layer transforms simple image data into sophisticated feature representations, achieving high spoof detection accuracy while maintaining a modular and manageable system architecture
3Measurement precision
If multiple convolution layers with different parameters are used to enhance detection accuracy, then spoof detection improves, but computational complexity increases
Solution Approach 1:
The patent applies partial action by using multiple convolution blocks with varying degrees of complexity. Not all blocks use the same maximum complexity; some use smaller kernels for quick feature extraction while others use larger kernels for deeper analysis. This selective application of computational resources achieves high detection accuracy while reducing overall energy consumption compared to uniformly complex processing
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and apparatus with spoofing consideration is provided. The method includes implementing convolution block(s) of a machine learning model that determines whether biometric information in an input image is spoofed, including generating a feature map including channels for an input feature map for the input image using convolution layers of a convolution block of the convolution block(s), in response to a total number of input channels of the convolution block and a total number of output channels of the convolution block being different, matching the total number of input channels of the convolution block and the total number of output channels of the convolution block by adding a zero-padding channel to the input feature map using a skip connection structure, and generating output data for determining whether the biometric information is spoofed, dependent on the generated feature map and a result of the skip connection structure.