Image Forensics—Analysis of Manipulated Images

With the emergence of low-cost internet connections, there has been significant growth in internet usage, and people are using social media to a larger extent. Social media forms a major part in human-to-human communication in the twenty-first century. Enormous number of images are shared and circulated on the social media platforms. Tampering and editing of the images are done even though internet and social media platforms like Facebook, Instagram, and Twitter connect people all over the world; evils in a society corrupt them. Small errors in information or manipulated images can wreak havoc and have unintended repercussions. Altered photos are a major source of misinformation/fake news and are frequently used maliciously, such as for mob provocation. Cases of similar incidents have grown during the previous decade. So, there should be one system which will filter this manipulated image before it leads to some major destruction. In the system mentioned in this paper, use of Error Level Analysis has been employed along with Convoluted Neural Network (CNN). The motive of this research work is to develop a system that will detect the manipulated image and verify its authenticity. By utilizing the machine learning algorithm ELA to detect changed photographs, the research helps reduce the dangers that edited images pose to society by keeping an eye on them. With a 98.5% accuracy rate, the Xception CNN model presented in this study performs well on the dataset. The dataset used is CASIA II, and it contains a total of 12,323 images and is furthermore divided into subsets. These folders consist of 7200 authentic images and 5123 tampered images.