Titre : | A CNN based architecture for forgery detection in administrative documents |
Auteurs : | Khadidja Maamouli, Auteur ; Abdelhamid Djeffal, Directeur de thèse |
Type de document : | Monographie imprimée |
Editeur : | Biskra [Algérie] : Faculté des Sciences Exactes et des Sciences de la Nature et de la Vie, Université Mohamed Khider, 2022 |
Langues: | Français |
Mots-clés: | CNN , Deep learning ,Forgery detection,Administrative scanned documents |
Résumé : |
In recent times, artificial intelligence (AI) becomes an invasion of scientific research fields, because it provides immediate and appropriate solutions, as it has touched many diverse areas of life. The administrative sector is one of the areas that enables (AI) to find solutions to its problems. This is what makes us sheding light on the phenomenon of administrative fraud through the forgery of administrative documents that the Algerian administration has been suffering from, as well as all administrations in the world. With the fast evolution of editing tools, fraudsters are continiously developping techniques that a simple naked eye inspection or even numeric solutions can not detect. For that reason, we attempt in this work to use machine learning techniques to learn fraudsters methods. Latest development in the field of artificial intelligence, especially deep learning, provides great solutions that help in predicting and detecting many of fraud features. This is why we explored in our project the possibility of deep learning (DL) to analyze scanned of administrative documents. In our study of the concepts of (DL), we dealt with Convolutional Neural Network (CNN) and artificial neural networks algorithms to build models that enable to automatically detect and classify documents into forged or authentic. We used Google colab platform to train our models. |
Sommaire : |
Contents Abstract ii Thanks v Dedication vi Table of Contents vii List of Figures ix List of Tables xii General introduction 1 1 Forgery of administrative documents 3 1.1 Introduction3 1.2 Administrative document 3 1.3 Administrative scanned document . 4 1.4 Forgery 4 1.5 Forgery types 4 1.5.1 Simple Forgery 5 1.5.2 Free Hand Simulation 5 1.5.3 Tracing5 1.5.4 Electronic Manipulation 5 1.6 Signs of Forgery 5 1.7 Forgery detection 6 1.8.1 Passive methods 6 viiCONTENTS 1.8.1.1 Copy-move forgery detection (CMFD) . 7 1.8.1.2 Splicing forgery detection 8 1.8.1.3 Imitation forgery detection 9 1.8.1.4 Image retouch detection 9 1.8.2 Active methods 10 1.8.2.1 Steganography 10 1.8.2.2 Cellular automata10 1.8.2.3 Watermarking11 1.9 Conclusion 11 2 Deep learning 12 2.1 Introduction 12 2.2 Artificial intelligence 12 2.2.1 Advantages12 2.2.2 Domains 13 2.2.3 The different fields of application of Artificial Intelligence 14 2.3 Machine learning 17 2.3.1 Artificial intelligence, Machine learning and Deep learning (AI , ML and DL) 17 2.3.2 Types of Machine learning algorithms18 2.3.3 Classification21 2.3.4 Regression21 2.4 Deep Learning 21 2.4.1 Artificial neural networks 22 2.4.1.1 Definition 23 2.4.2 Activation function 23 2.4.3 Type of neural networks 24 2.5 Convolutional neural networks 28 2.5.1 Definition 28 2.5.2 CNN layers types 28 2.5.3 CNN Architectures 30 2.6 Transfer learning 34 2.7 Related work (Dl and forgery detection) 35 viiiCONTENTS 2.8 Conclustion 36 3 Design of a deep learning architecture for forgery detection37 3.1 Introduction 37 3.2 General architecture 37 3.3 Detailed architecture 38 3.4 Conclusion 42 4 Implementation 43 4.1 Introduction 43 4.2 Frameworks, tools and libraries: 43 4.3 Implementation 46 4.3.1 Dataset preparation and preprocessing47 4.3.1.1 Functions and parameters of the used gaussin filter47 4.3.1.2 Splitting and preparing Dataset 48 4.3.2 Building our CNN Model 50 4.3.2.1 Import libraries and modules 50 4.3.2.2 Creating CNN Model 50 4.3.2.3 Model summary54 4.3.3 Training CNN Model 54 4.3.4 Testing our CNN Model:55 4.4 Obtained results . 56 4.4.1 Dataset 56 4.4.2 Presentation of the achieved performance 57 4.5 Comparison of our work with previous works 64 4.6 Conclusion 64 General conclusion 64 Bibliography 70 |
Disponibilité (1)
Cote | Support | Localisation | Statut |
---|---|---|---|
MINF/704 | Mémoire master | bibliothèque sciences exactes | Consultable |