Image Quality Assessment for Fake Biometric Detection

P. J. Suresh Babu

, K. Eswaramoorthy

, K. Sasikala

, P. Muthukumar

S. Sivamani

and T. Suresh Padmanabhan

Department of Electrical and Electronics Engineering, Saveetha Engineering College, Sriperumbadur, Chennai, Tamil

Nadu 602105, India

Department of Electrical and Electronics Engineering, Vels Institute of Science, Technology & Advanced Studies,

Chennai, Tamil Nadu – 600117, India

Department of Electrical and Electronics Engineering, Saveetha School of Engineering, Saveetha Institute of Medical and

Technical Sciences, Tamil Nadu - 602105, India

Keywords: Usefulness Expectations (UE), TV, Online Flow Experience (OFE), Internet Videos, Perceived Ease of Use

(PES), Perceived Behavioral Control (PBO).

Abstract: Detecting fake biometrics is crucial for ensuring the security and reliability of biometric authentication

systems. A recent paper proposes a method that extracts features from pre-processed images of the face and

fingerprint and compares them with those of a database image to obtain matching scores. The matching scores

then undergo a three-step process that includes normalization, generation of similarity scores, and fusion of

weighted scores. This ensures that the scores are on the same scale and comparable, allowing the system to

take advantage of the strengths of both biometric traits to increase accuracy and reduce false matches. The

weights for the scores obtained from the face and fingerprint traits are determined based on their individual

performance, and used to calculate a final score. The fusion of the face and fingerprint traits using a weighted

sum of scores technique has the potential to enhance the security of biometric authentication systems. The

proposed method ensures the correctness and dependability of the system by detecting fake biometrics and

preventing unauthorized access to sensitive information. Overall, this paper proposes an effective method to

improve the accuracy and reliability of biometric authentication systems by combining the strengths of two

biometric traits.

1 INTRODUCTION

Biometric authentication systems have become

increasingly popular in recent years due to their

ability to provide high levels of security and accuracy.

However, no single biometric trait is perfect and can

be affected by various factors, such as changes in

physical appearance or environmental conditions.

Therefore, combining multiple biometric traits can

improve the overall accuracy and robustness of the

system. The project aims to implement a biometric

authentication system that combines two biometric

traits, namely face and fingerprint, using a matching

score level architecture. The system extracts feature

from pre-processed images of the face and fingerprint

and compares them with those of a database image to

obtain matching scores. The individual scores

generated after matching are then passed through a

fusion module that consists of normalization,

generation of similarity scores, and fusion of

weighted scores. The fusion technique used in the

project is a weighted sum of scores technique, which

assigns weights to the individual scores based on their

relative importance. The final score obtained from the

fusion module is used to declare the person either

authenticated or unauthenticated. The proposed

system has the potential to improve the accuracy and

robustness of biometric authentication systems and

can be used in various applications such as access

control, secure transactions, and identity verification.

Image processing is a computer-based method of

modifying digital images using effective algorithms

to produce new images. The most popular software

for this is Adobe Photoshop, which is widely used for

processing digital photographs. Image processing is

used in various sectors, including face recognition,

medical imaging, and remote sensing. The process

involves taking a digital or analog image as input,

which is transformed into a physical picture using

Babu, P. J. S., Eswaramoorthy, K., Sasikala, K., Muthukumar, P., Sivamani, S. and Padmanabhan, T. S.

Image Quality Assessment for Fake Biometric Detection.

DOI: 10.5220/0013896500004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 3, pages

267-273

ISBN: 978-989-758-777-1

267

relevant technology. Digital photography saves

images as computer files, which are translated using

photography software to produce an actual image.

Image enhancement and correction are done using

specialized computer programs that use algorithms to

reduce signal distortion, clarify fuzzy photos, and

brighten dark images. While analog photography uses

chemicals to burn the picture onto film and requires

specialized training, digital photography is becoming

more popular due to its ease of use.

There are two categories of techniques in image

processing: analog and digital techniques. These can

process either using analog or visual techniques for

hard copies (e.g., printouts and photographs) as well.

These visual techniques are employed by image

analysts based on different principles of

interpretation. Image processing is not only limited to

a region that is analyzed but also the experience of the

analyst. Association is also an important technique in

image processing using visual methods. So what

analysts bring to image processing is the combination

of personal knowledge and collateral data. Digital

processing may be applied to processing of digital

images by computer. Because the raw data from the

imaging sensor on a satellite has shortcomings. In

order to overcome these imperfections and obtain the

original information, it must pass through several

processing stages. The three common stages that

should be addressed to handle every data type to be

used with digital methods are Pre-processing,

improvement and visualization and finally,

deconvolution. There are the five image processing

tasks. As follows:

• Visualization: Pay attention to intangible

objects.

• To improve the image, use image restoration

and sharpening.

• Search for the desired image using image

retrieval.

• Measures various things in a picture using a

pattern.

• Identify the things in a picture using image

recognition software.

Artificial Neural Networks and Representation

Learning are subsets of algorithms in the field of

deep learning (a subfield of machine learning) -

models and algorithms used to emulate human brain

and its natural processes. In computer vision, speech

recognition, natural language processing, audio

recognition, social network filtering, machine

translation, bioinformatics, drug design, medical

image analysis, material inspection and board game

programs, where they have produced results

comparable to and in some cases superior to human

experts. Deep learning models are vaguely inspired

in information processing and communication

patterns in biological nervous system and "deep belief

networks" have been fed data that is representative of

a wide range of noises, such as the chatter of telegraph

operators. Deep learning brings a higher recognition

rate than ever. For safety-critical use cases like self-

driving cars, this is a must-have to make sure

consumer electronics are reliable enough for

customers to take for granted. As deep learning’s

capabilities have improved in recent years, it has

begun to surpass humans in some tasks, like

classifying objects in images. IQA can be used to

detect image manipulation by analyzing changes in

image quality metrics. For example, if an image has

been manipulated to change the facial features of the

person, IQA can detect the changes in the facial

features and alert the system about the possibility of

a fake biometric image. IQA can also be used to

assess the authenticity of an image. Biometric images

are typically captured using specialized cameras and

have specific quality characteristics. By comparing

the quality of an image against a database of genuine

biometric images, IQA can detect anomalies and raise

alarms if the image appears to be fake.

IQA can also be used to identify specific image

tampering techniques that are commonly used to

create fake biometric images. By analyzing image

quality metrics, IQA can identify the presence of

artifacts and inconsistencies that are characteristic of

particular tampering techniques. IQA can also be

used to enhance the accuracy of biometric

authentication systems by identifying poor quality

biometric images. By removing poor quality images

from the database, IQA can improve the accuracy of

biometric matching and reduce the likelihood of false

positives and false negatives.

The problem statement for image quality

assessment for fake biometric detection is to develop

a reliable and accurate system that can differentiate

between genuine and fake biometric images. This

system should be able to assess the quality of the

image, detect any alterations or tampering, and

identify whether the biometric data captured is from

a real or fake source. The system should be able to

handle various types of biometric data, such as facial

images, fingerprints, iris scans, and voiceprints. The

goal is to improve the accuracy and reliability of

biometric systems for security and identification

purposes by ensuring that only genuine biometric data

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

268

is used for verification and authentication.

2 RELATED WORKS

However, it's important to note that technology

evolves rapidly, and there may have been

advancements in this field since then. Here are some

existing methods used for fake biometric detection:

Spoofing Detection: Spoofing refers to the use of fake

or artificial biometric traits to deceive a biometric

system. Various techniques have been developed to

detect spoofing attacks, such as liveness detection.

Liveness detection aims to verify that the biometric

being presented is from a live person and not a replica

or forgery. It may involve analyzing factors like skin

texture, blood flow, and thermal properties, or even

requesting specific actions from the user to prove

their liveness.

Presentation Attack Detection (PAD): PAD

techniques are designed to detect presentation attacks,

where an attacker presents a fake or manipulated

biometric sample to the system. PAD methods can

include analyzing the characteristics of the presented

biometric data to identify anomalies or

inconsistencies that indicate potential attacks. These

methods often involve analyzing the image quality,

texture, and other features to distinguish between

genuine and fake biometric data.

Multimodal Biometrics: Combining multiple

biometric modalities, such as face, iris, fingerprint,

voice, or behavioral characteristics, can enhance the

overall security and accuracy of a biometric system.

By using multiple biometric traits simultaneously, it

becomes more difficult for an attacker to spoof or

fake all of them convincingly. Multimodal biometric

systems can provide better resilience against spoofing

attacks.

Machine Learning and Artificial Intelligence:

Advanced machine learning algorithms, such as deep

learning, can be employed to train models that can

detect fake biometric data. These models can learn

patterns and features indicative of genuine or fake

biometric traits, allowing them to classify and

differentiate between them more accurately. By

continuously training the models with new data, they

can adapt and improve their detection capabilities

over time.

Database Comparison and Duplicate Detection:

Biometric systems often maintain databases of

enrolled biometric templates. By comparing newly

presented biometric samples against the existing

database, it becomes possible to detect potential

duplicates or inconsistencies that may indicate fake or

tampered data.

It's worth noting that the arms race between

attackers and system developers is ongoing, and new

spoofing techniques may emerge as technology

advances. Therefore, the field of fake biometric

detection continues to evolve, and researchers and

developers are constantly working to improve the

security and reliability of biometric systems.

2.1 Segmentation

The process of segmenting an image into various

areas or segments, each of which corresponds to a

different item or feature in the picture shown in figure

1 is known as "image quality analysis (IQA)".

Because it may be used to identify and isolate

particular areas of a picture that could include

irregularities or artifacts, segmentation can be helpful

in the identification of false biometrics. For instance,

segmentation may be used in a fingerprint

identification system to recognize and isolate certain

ridges or valleys in the fingerprint, which can be used

to detect modifications or fakes. In IQA,

segmentation is accomplished using a variety of

algorithms and methods, including thresholding,

clustering, and watershed segmentation. In order to

determine the borders between areas, these algorithms

examine the intensity or color gradients present

throughout the picture. After the picture has been

divided into segments, a number of metrics may be

generated for each segment to assess its authenticity

or quality. These metrics could include measurements

of form, size, homogeneity, contrast, and texture. As

a whole, segmentation are is a useful technique in

IQA for detecting false biometrics since it may assist

in separating paras of an image that might have

artifacts or anomalies, enabling more accurate and

consistent biometric authentication.

2.2 Feature Vector

A feature vector is a set of numerical values that

describe the characteristics or features of an image. In

image quality assessment (IQA) for fake biometric

detection, feature vectors are commonly used to

represent the biometric data in a way that is more

suitable for analysis and comparison. The process of

creating a feature vector typically involves extracting

relevant features or characteristics from the image,

such as texture, shape, and color, and quantifying

them as numerical values. The resulting feature

vector can then be used to compare the biometric data

with other samples and identify potential fakes or

Image Quality Assessment for Fake Biometric Detection

269

inconsistencies. There are several algorithms and

techniques used for feature extraction in IQA,

including wavelet transforms, principal component

analysis (PCA), and local binary patterns (LBP).

These algorithms work by analyzing different

aspects of the image, such as pixel values, edge

information, and texture patterns, to identify

relevant features. Once the feature vector has been

created, various metrics can be calculated to

determine the quality or authenticity of the biometric

data. These metrics may include measures of

similarity or distance between the feature vectors of

different samples, as well as measures of variability

and consistency within a single sample. Overall,

feature vectors are a valuable tool in IQA for fake

biometric detection, as they provide a way to

represent biometric data in a way that is suitable for

analysis and comparison, allowing for more accurate

and reliable biometric authentication.

Figure 1: Image Segmentation.

2.3 Limitations of the Existing System

This method is not very flexible because it is possible

to create duplicates of fingerprints or otherwise trick

the system. Currently, the system simply computes the

spatial domain, which is the order in which the data

are physically arranged. The spatial domain is merely

one type of data representation, and it may not be able

to offer a comprehensive picture of the data. The

paragraph recommends the application of the

Principal Component Analysis (PCA) technique to

enhance the biometric authentication system. By

spotting patterns and connections between variables,

PCA is a statistical approach used to simplify

complicated data sets. The system will be able to

calculate the co-variance and variance of the data and

locate more significant characteristics in the

biometric data by applying PCA. With the help of this

strategy, the system will be better able to identify and

authenticate people by giving a more accurate

depiction of the biometric data. Overall, the current

biometric authentication system has limitations and is

not very efficient, but a more advanced approach

using PCA can be used to improve the system's

accuracy and efficiency.

3 PROPOSED METHODOLOGY

3.1 Theoretcal Structure

Biometric systems refer to systems that identify

individuals on the basis of some of their physical or

behavioral characteristics and are usually employed

for security or access control. It’s high time for

organizations and companies to develop a system that

will utilise the facial and fingerprint as two varied

biometrics to validate an individual. It takes a photo

of a person’s face followed by a photo of his/her

fingerprint. These two biometric data sets are then

compared to biometric reference data and data from

the same two different biometric tests to determine

whether the stored biometric data corresponds to a

stored human's body. Facial recognition technology

does this by using unique landmarks on the person’s

face, such as the distance between their eyes or the

shape of their nose, to produce a mathematical

representation of the face. Fingerprint recognition

technology works instead by taking the unique ridge

and valley pattern of a person's fingertip and creating

a list of the minutiae points in it. 2 the same system

we used Fundamentals of Image Processing By using

both facial and fingerprint recognition technology the

system can produce more accurate and reliable

method. The multi-modal biometric feature

minimizes the occurrence of false positives and false

negatives that are associated with reliance on only

one modality of biometric feature. This method may

allow for a more reliable and/or secure access control,

especially in high security environments where

identification is essential.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

270

3.2 System Implementation

Fundamentals of Image Processing

Figure 2: Fundamentals of Image Processing.

The Figure 2 shows Fundamentals of Image Processing.

3.2.1 Image Acquisition

Image acquisition is the acquisition of a digital image.

To do so requires an image sensor and the capability

to digitize the signal produced by the sensor. The

sensor could be a monochrome or color TV camera

that produces an entire image of the problem domain

every 1/30 sec. The image sensor could also be a line

scan camera that produces a single image line at a

time. In this case, the object moved past the line. For

example: Digital Camera, Mobile Camera in figure 3.

Figure 3: Digital Camera and Mobile Camera.

4 RESULTS AND EVALUATION

4.1 The Project Setup

Start by clearly defining the problem you want to

solve. In this case, the problem is detecting fake

biometrics, which could involve analyzing images to

assess their quality. Collect a dataset of images that

includes both real and fake biometrics. You could

use publicly available datasets or create your own.

Depending on the quality of the data, you may need

to preprocess the images to remove noise, blur, or

other artifacts that could affect image quality. Define

metrics for evaluating the quality of the images. For

example, you could use measures like sharpness,

contrast, or texture features to assess the quality of the

images. Choose a model to assess image quality. This

could include traditional image processing techniques

or more advanced machine learning algorithms. Train

the model using the dataset you have collected. You

may need to fine-tune the model to achieve the best

performance. Evaluate the model's performance on a

test set of images that it has not seen before. This will

give you an idea of how well the model can generalize

to new data. Once you are satisfied with the

performance of the model, integrate it into your

biometric detection system. Deploy the system and

monitor its performance in real-world scenarios.

Continuously collect feedback from users and update

the model to improve its performance over time.

4.2 Dataset Used in this Fake Biometric

Detection

The NIST Biometric Image Software (NBIS) provides

datasets of fingerprint, face, and iris images that are

suitable for training and testing biometric systems.

The Cross-Match dataset includes both genuine and

fake fingerprint images shown in Figure 4. It is widely

used in research on fake biometric detection. The 1:1

Verification Competition dataset from the

International Conference on Biometrics (ICB)

includes a large number of face images, including

both genuine and fake images. The Mobile Biometry

(MOBIO) dataset includes images of faces,

fingerprints, and voices, with both genuine and fake

samples. The Mobi Face dataset contains face images

captured in uncontrolled environments and includes

both genuine and fake images. The Replay-Attack

dataset includes both genuine and fake face images

captured under various conditions to simulate

different attack scenarios. The MSU Mobile Face

Spoofing Database includes both genuine and fake

face images in Figure 5, captured using high-quality

Image Quality Assessment for Fake Biometric Detection

271

cameras and mobile devices.

Figure 4: Fingerprint Dataset.

Figure 5: Face Dataset.

5 CONCLUSIONS

Fake biometric images can be generated using various

techniques, such as printing, replay, or synthetic

generation. These techniques can create images that

are similar enough to real biometric images to be

accepted by traditional biometric verification

techniques, making them difficult to detect. Image

quality assessment can help identify fake biometric

images by analyzing various image features such as

sharpness, contrast, and noise. By detecting anomalies

in these features, image quality assessment algorithms

can flag potentially fake images for further

investigation. For example, a fake biometric image

created using printing may exhibit different

characteristics than a real biometric image captured

using a biometric sensor. However, it is important to

note that image quality assessment is not foolproof

and can be bypassed by advanced attacks. Attackers

can try to mimic the image characteristics of real

biometric images, making them harder to detect.

Therefore, it is important to continuously improve and

update image quality assessment algorithms to stay

ahead of evolving attack techniques. Overall, image

quality assessment is a valuable tool in the fight

against fake biometrics and can help improve the

security and reliability of biometric systems. By

identifying and rejecting fake biometric images,

image quality assessment can help ensure that only

genuine users are granted access to protected

resources, improving the overall security posture of

the system. While current image quality assessment

algorithms can detect basic anomalies in image

features, there is room for improvement. Future work

could focus on developing more sophisticated

algorithms that can detect subtle differences between

real and fake biometric images. Machine learning

techniques, such as deep learning, have shown

promise in improving the accuracy of image quality

assessment. Future work could explore the use of

machine learning to enhance the performance of

image quality assessment algorithms. Image quality

assessment algorithms may perform differently

depending on factors such as the type of biometric

modality, lighting conditions, and the quality of the

biometric sensor. Future work could evaluate the

performance of image quality assessment algorithms

under a range of scenarios to identify areas for

improvement. Image quality assessment algorithms

typically rely on metrics such as sharpness, contrast,

and noise to identify anomalies in biometric images.

However, there may be other metrics that could be

useful in detecting fake biometric images. Future

work could focus on identifying and developing new

metrics that could be incorporated into image quality

assessment algorithms. As noted earlier, image

quality assessment is not foolproof and can be

bypassed by advanced attacks. Future work could

focus on developing complementary techniques for

fake biometric detection, such as liveness detection,

which can help detect attacks that bypass image

quality assessment.

REFERENCES

A. Peña, A. Morales, I. Serna, J. Fierrez, and A. Lapedriza,

"Facial expressions as a vulnerability in face

recognition," in 2021 IEEE International Conference on

Image Processing (ICIP), pp. 2988-2992, IEEE,

September 2021.

A. Brömme, C. Busch, N. Damer, A. Dantcheva, M.

Gomez-Barrero, K. Raja, C. Rathgeb, A. Sequeira, and

A. Uhl (Eds.), "Image quality assessment on identity

documents," in BIOSIG 2021, Lecture Notes in

Informatics (LNI), Gesellschaft für Informatik, Bonn,

2021.

A. George and S. Marcel, "On the effectiveness of vision

transformers for zero-shot face anti-spoofing," in 2021

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

272

IEEE International Joint Conference on Biometrics

(IJCB), pp. 1-8, IEEE, 2021.

J. Cui, P. Zhang, S. Li, L. Zheng, C. Bao, J. Xia, and X. Li,

"Multitask identity-aware image steganography via

minimax optimization," IEEE Transactions on Image

Processing, vol. 30, pp. 8567-8579, 2021.

M. Raif, A. E. Rharras, A. Chehri, and R. Saadane,

"Metamorphic testing for edge real-time face

recognition and intrusion detection solution," in 2022

IEEE 96th Vehicular Technology Conference

(VTC2022-Fall), pp. 1-5, IEEE, 2022.

P. Anthony, B. Ay, and G. Aydin, "A review of face anti-

spoofing methods for face recognition systems," in

2021 International Conference on Innovations in

Intelligent Systems and Applications (INISTA), pp. 1-

9, IEEE, 2021.

X. Zhu, H. Wang, H. Fei, Z. Lei, and S. Z. Li, "Face forgery

detection by 3D decomposition," in Proceedings of the

IEEE/CVF Conference on Computer Vision and

Pattern Recognition, pp. 2929-2939, 2021.

Y. Wang, X. Song, T. Xu, Z. Feng, and X.-J. Wu, "From

RGB to depth: Domain transfer network for face anti-

spoofing," IEEE Transactions on Information Forensics

and Security, vol. 16, pp. 4280-4290, 2021.

Z. Li, H. Li, X. Luo, Y. Hu, K.-Y. Lam, and A. C. Kot,

"Asymmetric modality translation for face presentation

attack detection," IEEE Transactions on Multimedia,

vol. 25, pp. 62-76, 2021.

Z. Li, R. Cai, H. Li, K.-Y. Lam, Y. Hu, and A. C. Kot, "One-

class knowledge distillation for face presentation attack

detection," IEEE Transactions on Information

Forensics and Security, vol. 17, pp. 2137-2150, May

26, 2022.

Image Quality Assessment for Fake Biometric Detection

273