Deep learning-based image forgery detection system
Helina Rajini Suresh, M. Shanmuganathan, T. Senthilkumar, B. S. Vidhyasagar
Abstract
Despite the fact that there are more complex ways of forgery being developed all the time, image forgery detection continues to play an essential part in the field of digital forensics. The problem of counterfeit photographs is a worldwide problem today that is mostly distributed via social networking sites. The ability to identify phoney pictures eliminates the possibility that fraudulent photographs may be used to trick or damage other people. Within the scope of this research, we investigate the deep learning technique to image forgery detection. The proposed model implemented by python language uses input images in batches and a convolutional neural network (CNN) using ResNet50v2 architecture and YOLO weights. We analysed the CASIA v1 and CASIA v2 benchmark datasets. For the purposes of training, we used 80% of the data, and the remaining 20% was used for testing. 85% accuracy obtained for the dataset.