International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 09 Issue: 06 | June 2022
p-ISSN: 2395-0072
www.irjet.net
Image Forgery / Tampering Detection Using Deep Learning and Cloud Misbah Shaikh 1, Dr. Dipak Patil 2 1Student,
Department of Computer Engineering, Gokhale Education Society's, R. H. Sapat College of Engineering, Management Studies and Research, Nashik, India 2 Professor ,Department of Computer Engineering, Gokhale Education Society's, R. H. Sapat College of Engineering, Management Studies and Research, Nashik, India ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Cybercrime has become more prevalent in recent
more difficult to detect. Motives for generating forged photographs might range from financial gain to spreading rumors or making false claims in one's favor.
years. With modern photo editing tools as widely available as ever, it has been demonstrated that creating phony papers is incredibly simple [1]. With the help of this tool, which offers tools for doing so, documents can be scanned and forged in minutes. While photo editing software is convenient and widely available, there are also deft methods for investigating these transformed documents. This study presents a framework for investigating digitally modified documents as well as a way of distinguishing between an original document and a digitally morphing document. we created a web application to detect digitally modified photos. This method has more than 95.0 percent accuracy and has proven to be efficient and useful. Recent work on forgery detection using neural networks has proven to be very effective in detecting image forgery additionally we are using Azure Form recognizer service to read data from documents and verify it on the server, this dual approach makes the system robust and very accurate. Deep Learning methods are capable of extracting complex features in an image, resulting in increased accuracy. In contrast to traditional methods of forgery detection, a deep learning model automatically builds the required features, and as a result, it has emerged as a new area of study in image forgery.
Deep learning is revolutionizing the field of computer vision, which is already expanding [3]. A CNN is a cutting-edge deep learning technology that learns high-level characteristics from a vast collection of labelled images. Ink analysis in document image processing allows for the determination of ink age and forgery, as well as the identification of the pen or writer. Ink spectral information in hyperspectral document pictures provides essential information about the underlying material, assisting in the identification and discrimination of inks based on their distinct spectral signatures, even though they are the same hue.
1.1
A System Based on Intrinsic Properties for Fraudulent Document Detection, the authors offer an automatic forgery detection system based on the intrinsic features of the document at the character level in this study [4]. This method is based on outlier character detection in a categorical feature space on the one hand, and strictly similar character detection on the other. As a result, a feature set is computed for each character. The character is then classed as authentic or fraudulent depending on the distance between characters of the same class. Local Binary Patterns for Detecting Document Forgery.
Key Words: morphed document, Azure form recognizer, CNN, Deep learning, neural network.
1. INTRODUCTION
The authors of this study [5] describe a classification-based technique to forgery detection. The author employs uniform Local Binary Patterns (LBP) to capture discriminant textural characteristics seen in fabricated regions. Furthermore, the author combines numerous descriptors from nearby locations to simulate contextual information. The results of patch classification using Support Vector Machines (SVM) reveal that people can recognize numerous types of forgeries in a wide range of document categories.
With the tremendous technological improvement that has boosted the progress of every industry imaginable, one of which is security, it has also become easier to break it [2]. Legal documents can be stolen and faked, but criminal evidence, such as photographs and security footage, can also be easily tampered with. One may believe that checking IDs at the front gate is sufficient for an institution, but they do not comprehend how simple it is for a criminal to obtain false IDs. Even for inexperienced crooks, posing as someone else in public is a simple task. As previously stated, photo editing tools are not only easily available but also incredibly user-friendly. Even if you've never used picture editing software before, you can master fundamental photo editing techniques in a few hours. Photo editing is no longer particularly advanced, and counterfeiting has grown even
© 2022, IRJET
|
Impact Factor value: 7.529
LITERATURE REVIEW
This study [6] introduces a new method for detecting fake documents. The proposed method is based on network science methodologies for assessing ink spectrums of documents to determine whether they were counterfeited or fabricated. Laser-Induced Breakdown Spectroscopy is utilized to extract the spectrums of the original and
|
ISO 9001:2008 Certified Journal
|
Page 2105