A Tale of a Deep Learning Approach to Image Forgery Detection

Social media has become a part and parcel in people's lives in modern days. Due to the high availability of cameras and presence of various image editing tools in mobiles, it has become much easier to forge fake images. Such false and fake images can and are being used for slandering reputation...

Full description

Saved in:
Bibliographic Details
Published in2018 5th International Conference on Networking, Systems and Security (NSysS) pp. 1 - 9
Main Authors Majumder, Md. Taksir Hasan, Alim Al Islam, A. B. M.
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.12.2018
Subjects
Convolutional Neural Network
Deep learning
Discrete cosine transforms
Feature extraction
Forgery
Image Forgery Classification
Image Splicing
Media
Signal processing
Splicing
Online AccessGet full text

Cover

More Information
Summary:Social media has become a part and parcel in people's lives in modern days. Due to the high availability of cameras and presence of various image editing tools in mobiles, it has become much easier to forge fake images. Such false and fake images can and are being used for slandering reputation of popular figures and in dirty political moves. The problem of detecting and localizing such images is known as image forgery detection. Due to how small the edited regions are compared to the full image size, deep learning without manual feature engineering approaches are quite difficult to apply. In this paper, we focus on binary classification of forged or authentic images, and we highlight how deep learning models should be leveraged for good performance in forged image classifications.
ISBN:9781728113241
1728113245
DOI:10.1109/NSysS.2018.8631389