Scholarlink

This paper investigates image forgery, which is a significant problem in the field of digital images. In recent years, many researchers have focused on Convolutional Neural Networks (CNNs) to detect image forgeries. However, there is still a lack of a comprehensive analysis of the effectiveness of these methods. In this study, we propose a novel approach that this work analyse the implementation of CNN based algorithm for Copy-Move Forgery Detection (CMFD). The study aims to explore and compare the efficacy of the error-level analysis approach and CNN, in detecting fake images. The paper highlights the significance of addressing the limitations of Error Level Analysis (ELA) when combined with CNNs, particularly in the context of Copy-Move Forgery Detection (CMFD). This paper demonstrates that the ELA-CNN model achieves an overall training and validation accuracy of 92.00% and 87.60%, respectively, using VGG16 Architecture, which shows the efficiency and robustness of the technique in identifying forgeries.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Convolutional Neural Network (CNN) for Image Forgery Detection

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Gnanasegaran Maheswary,
  • Ruzelita Ngadiran,
  • Iszaidy Ismail

摘要

This paper investigates image forgery, which is a significant problem in the field of digital images. In recent years, many researchers have focused on Convolutional Neural Networks (CNNs) to detect image forgeries. However, there is still a lack of a comprehensive analysis of the effectiveness of these methods. In this study, we propose a novel approach that this work analyse the implementation of CNN based algorithm for Copy-Move Forgery Detection (CMFD). The study aims to explore and compare the efficacy of the error-level analysis approach and CNN, in detecting fake images. The paper highlights the significance of addressing the limitations of Error Level Analysis (ELA) when combined with CNNs, particularly in the context of Copy-Move Forgery Detection (CMFD). This paper demonstrates that the ELA-CNN model achieves an overall training and validation accuracy of 92.00% and 87.60%, respectively, using VGG16 Architecture, which shows the efficiency and robustness of the technique in identifying forgeries.