Comparative study of presentation attack detection methods for finger vein recognition
Biometrics refers to the physiological and behavioural characteristics uniquely possessed by individuals. Physiological biometrics are characteristics that can be measured from the human body such as palm print, fingerprint, iris, and finger vein. Meanwhile, behavioural biometrics are measurements o...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2023
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Biometrics refers to the physiological and behavioural characteristics uniquely possessed by individuals. Physiological biometrics are characteristics that can be measured from the human body such as palm print, fingerprint, iris, and finger vein. Meanwhile, behavioural biometrics are measurements of patterns of humans such as gait, signature, and gesture. Among the different biometrics in use today, finger vein biometrics have been gaining popularity and are widely used for user authentication especially in financial and access control applications. Generally, finger vein recognition is considered a secure biometric, as the finger vein pattern resides in the human finger, making it difficult to steal. Unfortunately, there are attack attempts known as presentation attacks to circumvent the system by presenting fake finger vein images with the intent of spoofing the finger vein sensor or reader. A presentation attack is discovered at the sensor level during the image acquisition step. After obtaining the finger vein images, the attackers can duplicate them, pre-process them, and then print them on various types of paper. These forgeries are then presented to the finger vein sensor. This type of presentation attack, according to the literature, can have a false acceptance rate of up to 86%. This research aims to develop different presentation attack detection methods for finger vein recognition. This study proposes three presentation attack detection methods: 1) Local Directional Pattern (LDP), which focuses on the edge response of the finger vein, 2) Multi-scale Histogram of Oriented Gradient (MHOG), which analyses gradient intensity of the image, and 3) Block-wise Variance-based Image Quality Assessment (BV-IQA), which evaluates the discrimination of noisiness and blurriness information in the finger vein images. The first two methods are texturebased approaches, whereas the third is an image quality assessment approach. Experiments conducted on the SCUT and VERA benchmark datasets validate the efficacy of the proposed methods with the BV-IQA achieving the lowest ACER of 0.04%. |
|---|