An Efficient Image Forgery Detection Framework using Transfer Learning Models

In recent years, digital images have become common in modern life in various applications. This leads to be attractive to attackers to modify them for malicious activities. The process of modification of digital images is known as image forgery. There is a wide range of software and hardware tools available for modifying and producing them. It is of utmost importance to understand how authentic and forged images can be detected and classified since digital image forgery can be extremely damaging. Since the advent of digital photography, there have been new and creative ways to forge images. When images are altered by forgeries, it affects how the viewer perceives the scenario that is being shown, which can have negative effects if done maliciously. Therefore, there is a serious need to provide a method to detect the modification of digital images. Thus, this paper provides an efficient image forgery framework using transfer learning models for the effective detection of image forgeries. From the results, the proposed method proved to be most accurate in detecting image forgeries on a dataset containing a mixture of 220 authentic and forgeries images, these models as RaseNet50, VGGI6, Xception, Densenet201, MobileNet, and NasNetMobile. The RaseNet50 gives the highest accuracy of 96%.to detect image forgeries in a real-time manner.