A Machine Learning Based Classifier for Automated Detection of Image Forgery

With social media making its presence felt wide and far, its consequences are also far reaching at least in the context of imagery. The amount of information that can be shared by images is enormous as compared to test data. However, there remains a chance of fake and forged image data that can be circulated which can result in disastrous consequences for individuals, firms or communities at large. With the advancements in image editing tools, it is practically impossible to detect fake or forged images by the naked human eye. Moreover, the number of images shared specifically on social media platforms is so large that human intervention is practically infeasible. In this paper, image forgery detection has been carried out using artificial neural networks and image processing techniques. The configuration of the neural networks is the ada-boost network. The performance index is the classification accuracy. It has been shown that the proposed technique achieves higher classification accuracy compared to previously existing methods for the same dataset.