Image Forgery Detection using Deep Learning Model

The act of tampering with or altering an image is known as image forgery. This image manipulation may appear to be harmless. When an image is used in legal proceedings or forensic investigations, however, it must be unaltered to maintain the integrity of the evidence. Manipulation of an image can be done in a variety of ways. There are numerous ways to predict image forgery, just as there are numerous ways to predict forgery. However, the problem with those methods is that they are so advanced that they are inaccessible to the general public. To resolve this issue, this study comes up with the idea of the developing of a Deep Learning (DL) model. For this purpose, a dataset consisting of both original and forged images is collected from a database named CASIA. This dataset is then preprocessed. The preprocessing techniques include image resizing and denoising. The processed images are then used to train the DL model. The working of the model is then tested using the accuracy and loss as the main parameters. In the end, it is found that the model is capable of producing an accuracy greater than 95% which is greater than many other DL models. Just like the accuracy, the loss value is also found to be very low that it can be neglected. This makes the model more efficient. In the future the model can be deployed into a website, making it useful for the common people.