Machine Learning Techniques for Image Manipulation Detection: A Review and Analysis

Abstract

Low-cost modified or tampered image enhancement processes and advanced multimedia technologies are becoming easily obtainable as image editing methods, different editing software, and image altering tools are becoming advanced. These manipulated multimedia images or videos can be used to fool, attract or mislead the public or readers, malign a person’s personality, business, political opinions and can affect in criminal inquiry. Most of the research have been undertaken this research work to find the solution to the image manipulation using the deep learning (DL) methodologies that deal with the problem of determining and distinguishing the tempered regions in real and fake images. In this work, we present a study of existing machine as well deep learning-based image manipulation detection approaches. The survey found that the researchers have paid more attention to image content while paying less attention to tempering artifacts and other image features. The primary focus of this research is on ML-based solutions of image manipulation detection. The presented study examines various techniques for determining whether image is genuine or tempered. Besides, the review paper presents a comparative study of various effective approaches in image manipulation field. A brief summary of various datasets used for image manipulation detection is also presented. Finally, the research challenges that open for further research work are listed. Although, lot of work has been done to find the manipulations in the images and videos but these approaches still fall short in some areas.