International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 10 Issue: 06 | June 2023
p-ISSN: 2395-0072
www.irjet.net
Image Forgery Detection Using Deep Neural Network Dr. N P Nethravathi1, Bylla Danny Austin2, Dadireddy Sai Praneeth Reddy3, Grandhi Venkata Naga Satya Pavan Kumar4, Guduru Karthik Raju5 1Professor, School of Computer Science and Engineering, REVA University, Karnataka, India 2Student, School of Computer Science and Engineering, REVA University, Karnataka, India 3Student, School of Computer Science and Engineering, REVA University, Karnataka, India
4Student, School of Computer Science and Engineering, REVA University, Karnataka, India 5Student, School of Computer Science and Engineering, REVA University, Karnataka, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The detection of fake images is crucial to
As a result, the identification of modified photographs has become a key area of research in the field of image forgery detection. These methods can be broadly divided into two categories: passive ones that don't require knowledge of the original image and active ones that involve adding data to the image to aid in subsequent authentication. Finally, complete content and organizational editing before formatting. Please take note of the following items when proofreading spelling and grammar.
maintain the credibility of digital content, especially in the current era of digital media and social networks. Image forgery has become increasingly common and sophisticated, posing a serious threat to the authenticity and validity of digital content. This paper presents a deep learning-based approach to image forgery detection, specifically using Error Level Analysis (ELA) with Convolutional Neural Networks (CNNs) and a pre-trained VGG-16 model. The study compares the performance of the two models and provides an in-depth analysis of the results. The experiments show that the ELACNN model achieves a remarkable accuracy rate of 99.87% and correctly identifies 99% of invisible images, while the VGG16 model achieves a lower accuracy rate of 97.93% and a 75.87% validation rate. The research highlights the significance of using deep learning techniques in image forgery detection and explores the implications of the findings. The paper also discusses the limitations of the study and future enhancements that could be made to improve the precision and generalization skills of image forgery detection algorithms. This research contributes to the field of image forgery detection by providing a comprehensive comparison of deep learning-based algorithms and their effectiveness in identifying fake images. The findings of this study can be utilized to develop precise and effective image forgery detection tools to maintain the integrity of digital content and mitigate the negative consequences of picture alteration.
1.2 Significance and motivation for research in this field With the growth of digital media, social networks, and the extensive transmission of information online, image forgery detection has become very important. Forgeries can have serious repercussions in a variety of industries, such as journalism, law enforcement, and social media, as modified photographs can mislead the public, damage people's reputations, or support false narratives.
Key Words: Image Forgery, Machine Learning, CNN, Deep Learning, Neural Network, Error-level Analysis.
The integrity of digital content can be maintained, and the negative consequences of picture alteration can be mitigated by the development of precise and effective image forgery detection tools. Additionally, in order to maintain their effectiveness, detection systems must advance along with counterfeit techniques. Particularly when it comes to enhancing the precision and generalisation skills of image forgery detection algorithms, deep learning techniques have shown considerable promise.
1. INTRODUCTION
1.3 Scope of the paper and its contributions
1.1 Background of image forgery detection
To increase the precision of image forgery detection, this research focuses on using deep learning approach, notably CNNs and a pre-trained VGG-16 model. We give an overview of current methodologies and their shortcomings before delving deeply into our suggested strategies employing Error Level Analysis (ELA) with CNNs and a trained VGG-16 model. Our test findings show that whereas the VGG-16 model produces a 97.93% accuracy and a 75.87% validation rate, the ELA-CNN model achieves a 99.87% accuracy and correctly recognizes 99% of unknown images.
Image manipulation has grown to the point where it is both common and sophisticated because of the quick growth of digital image processing tools. The purposeful change of an image's content to deceive or send false information is known as image forging, and it poses a severe danger to the validity and authenticity of digital content. A variety of methods, such as copy-move, image retouching, and splicing, have been used to create forgeries that are challenging to detect from visual inspection alone.
© 2023, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 1095