CHAPTER ONE

INTRODUCTION

1. 1. Background of the Study

Digital image manipulation has become increasingly common in recent years for a variety of reasons, including digital image fraud, which includes the replacement of fingerprints to doctor a fingerprint image (Singh et al, 2021). Digital image manipulation is simple with the help of readily available digital processing tools such as Photoshop (Liu, 2021).

Digital image manipulation includes the following: 1) copy-paste manipulation, which includes an uncompressed inserted region but saves the composite image in the Joint Photographic Expert Group (JPEG) format; 2) copy-paste manipulation, which includes a compressed inserted region and saves the composite image in JPEG format; and 3) inpainting manipulation of JPEG images (Liu et al., 2022). Face images, fingerprint images, and iris images, among other biometric modalities, are stored in biometric databases (Gayathri & Malathy, 2021). The majority of these modalities are captured digitally and saved in JPEG format (DaCunha et al., 2021). These databases are frequently kept as backups, whereas a database of derived features is kept for operational purposes. However, it is not impossible for insider attackers to gain access to the backup copy of the database and manipulate it.

The majority of research is conducted with the use of face and fingerprint biometric modalities for verification, identification, and as evidence in court. People may tamper with such biometric modalities maliciously, either in the database or before presenting them as evidence in court. Several but distinct research efforts have been made on image tampering detection, face recognition in forensic scenarios, and other topics, but to the best of the  knowledge, only a few studies have been conducted on verifying the integrity, authenticity, and reliability of biometric modalities using a deep learning framework.

This study proposes the use of a deep learning model to aid in the detection of tampered biometric fingerprint images in biometric databases, inspired by digital tampering of stored biometric fingerprints in databases and the possibility of intentional or accidental use of particular fingerprints instead of another fingerprint. Furthermore, the study proposes that deep learning can be used to detect the tampered region of tampered biometric fingerprints.

1.2 Statement of the Problem

One of the biggest issues that biometric systems face today is the increasing threat of malicious operations. To overcome biometric systems, the majority of malicious actors use a common sort of presentation attack known as "spoofing" (Uliyan et al, 2020). The primary purpose of a presentation attack is to impersonate target victims with the appropriate authorization. It occurs when intermediate spoofing forgers steal victims' fingerprints and alter them with legal content in order to fool fingerprint recognition systems.

As a result, a method for detecting biometric data tampering as well as identifying the region where the data has been tampered with is required. The detection and localization of fingerprint tampering using a deep learning model, has been proposed in this study.

1.3    Aim and Objectives of the Study

The aim of this is study is to create a system that can detect tampered fingerprints images and identify the region tampered with in the image.

The objectives of this study include:

1. To develop a deep neural network model that detects tampered fingerprints and locate the region that has been tampered in the fingerprint image.
2. Train and test the model developed in other to optimize performance
3. Evaluate the functionality of the models developed.

1.4     Significance of the Study

Fingerprint identification and verification is a popular biometric approach because of its ease of use, uniqueness, and long-term durability. Fingerprints play an important role in public safety and criminal investigations, including forensic inquiry, law enforcement, tax access, and public security. In various applications, such as video surveillance, biometric identification, and face indexing in social media, failure to prevent fingerprint image tampering on devices may compromise personal information (Khan et al., 2020). As a result, countermeasure measures must be used and integrated into biometric-based systems to prevent such occurrences.

As a result, the significance of this study is to develop a system that can tampered images of fingerprints  using a deep learning model called a convolutional neural network, allowing for a more reliable biometric verification and authentication system.

1.5 Scope of the study

The system will be implemented using deep learning framework to train images of real and tampered fingerprints images; after which the trained model will be tested on test fingerprint image for evaluation.

1.6 `Limitations of the study

The Limitations of the study is that the system implemented cannot detect other biometric data like face or signature.

1. 1. 1. Definition of Terms
1. Fingerprint: A fingerprint is an impression left by the friction ridges of a human finger. The recovery of partial fingerprints from a crime scene is an important method of forensic science
2. Digital image manipulation: Digital image manipulation is the process of digitally editing an image. It is often referred to as “photoshopping,” in reference to Adobe’s popular photo-editing software called photoshop

3. Spoofing: In the context of information security, and especially network security, a spoofing attack is a situation in which a person or program successfully identifies as another by falsifying data, to gain an illegitimate advantage.

4. JPEG: JPEG stands for “Joint Photographic Experts Group”. It’s a standard image format for containing lossy and compressed image data. Despite the huge reduction in file size JPEG images maintain reasonable image quality. This unique compression feature allows JPEG files to be used widely on the Internet, Computers, and Mobile Devices. 

5.  Biometric Systems: the automated recognition of individuals based on their behavioral and biological characteristic

6. Algorithm: An algorithm is a process or set of rules to be followed in calculations or other problem solving operations, especially by a computer.

7   Deep learning: Deep learning is a subset of a larger family of machine learning techniques based on representation learning and artificial neural networks. There are three types of learning: supervised, semi-supervised, and unsupervised.

8   Feature extraction: Process of building informative, no redundant derived values from an initial set of measured data to facilitate learning, generalization and interpretation.

9   Biometric Tampering: Biometric tampering is the process of subjecting biometric data to many malicious attacks which can be performed by various forms of threats. Malicious treats on a biometric data are a security concern and degrade the system's performances.

10 Spoofing: spoofing is when fraudsters pretend to be someone or something else to win a person’s trust. The motivation is usually to gain access to systems, steal data, steal money, or spread malware.

11. Presentation Attack: A Presentation attack is a technique of deceiving a biometric system using a specific tool.