Image Forgery Detection

Abstract

Images shared online have a high likelihood of being altered, and further global alterations like compression, resizing, or filtering mask the potential change. Many restrictions are placed on forgery detection systems by such manipulations. Image forgery detection is the fundamental solution to many issues, particularly social issues like those on Facebook and legal issues. The most frequent form of image fraud is called a copy-move forgery, where a portion of the original image is copied and pasted in a different spot within the same image. Because the duplicated portions' attributes are similar to those of the original image's components, this type of picture counterfeiting is simpler to carry out but more challenging to detect. The method for spotting copy-move forgeries described in this study is based on processing blocks into features and then extracting those features from the blocks' transforms. A Convolutional Neural Network (CNN) is another tool for detecting forgeries Serial pairings of convolution and pooling layers are employed to conduct feature extraction. Original and changed images are then categorised using transforms and without transformations. We use the CASIA2 dataset, which has 4795 images, of which 1701 are authentic and 3274 are forged. The accuracy of our proposed model is 97.7%. This improved the detection process's overall processing effectiveness and allowed it to fulfill real-time processing demands..