Image Forgery Detection Using Convolutional Neural Networks

• Python
• Machine Learning

In recent years, with the prevalence of cameras, taking pictures has become more and more popular. Images are essential to our daily life as they contain a wealth of information and often need to be enhanced to obtain additional information. The detection of fake images is crucial to 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. Various tools are available to improve image quality. Nevertheless, they are also commonly used for fake images, leading to the spread of misinformation. This has increased the severity and frequency of image forgery, a major concern today. Many traditional techniques have been developed over time to detect fake images. Convolutional Neural Networks (CNNs) have gained a lot of attention in recent years, and CNNs are also influencing the field of image forgery detection. However, most image forgery techniques that exist in the CNN-based literature are limited to detecting specific types of forgery (image splicing or copy-transfer). Therefore, there is a need for techniques that can efficiently and accurately detect the presence of invisible counterfeits in images. In this article, we present a robust deep learning-based system (CNN) that uses mantra-net to identify image fakes. ManTraNet is end-to-end convolutional neural network consisting of two sub-networks. One for extracting features associated with tamper evidence and the other for detecting local anomalies between features. A graphical user interface (GUI) was created for detecting digitally processed images. The method has an accuracy of 98.4% and has been proven to be efficient and practical.