Image-based academic certificate documents forgery detection using deep  learning approach

Tegbaru, Nibret

dc.contributor.author	Tegbaru, Nibret
dc.date.accessioned	2023-06-19T07:23:24Z
dc.date.available	2023-06-19T07:23:24Z
dc.date.issued	2023
dc.identifier.uri	http://ir.bdu.edu.et/handle/123456789/15394
dc.description.abstract	Digital image forensics (DIF) is produced by altering or modifying the original image with different techniques. Forged academic certificate documents are formed by copy-move, splicing, and resampling techniques. Deep learning is the basic class of machine learning algorithms to imitate the working of computers as humans. It uses multiple layers to extract effective features. Convolutional neural network (CNN) is part of a deep learning approach and addresses the point of feature extraction of images. However, the enhancement of technology has the challenge of detecting the authenticity of academic certificate documents, and difficult to find the forgeries by the human naked eye. The most common form of image manipulation technique is region duplication by copy and move forgery where a portion of the image is copied and pasted to another portion in the same digital image. To investigate such forensic analysis, various techniques and methods have been developed by scholars. The objective of this research work is to develop academic certificate document forgery detection by using a deep learning approach. For the detection and classification of forged academic documents, a deep learning approach is proposed and algorithms are used for image preprocessing. For image segmentation, OSTU‟s Threshold algorithm, k nearest-neighbor (KNN), and water shade algorithm are experimented with, in which the latter is used for extracting a region of interest. Feature extraction is performed by CNN feature extractor with transfer learning models (such as VGG16, VGG19, MobileNetV2, and ResNet50). Finally, CNN with Support vector machine (SVM) and CNN with fine-tuning would be used for classification. An experimental result shows that CNN with fine-tuning performs better than CNN with SVM with an accuracy of 84.44%. The challenge in this research work is collecting a large number of forgery academic certificate documents in organizations; hence we recommend the need to prepare standardized datasets to enhance the performance of the forgery detection model. Keywords: Certificate document forgery, Deep Learning Approach, Detection of certificate document forgery, Support vector machine.	en_US
dc.language.iso	en_US	en_US
dc.subject	Computing	en_US
dc.title	Image-based academic certificate documents forgery detection using deep learning approach	en_US
dc.type	Thesis	en_US