Image Forgery Detection Using Deep Learning

Images playa significant role in our life. With the growing Images are playing a pivotal role in our daily life. By the growing era images are becoming more and more popular as it contains lot of informatic content as well as memorable too. The problem arises from here, as the images are becoming popular day by day there are also increase in counterfeiting of the images or we can say that image forgery. It leads everyone to believe the content that's been forged. Not everyone is able to detect whether the image is forged or not. There are several applications that help people forge the image. This is a serious problem that has been arising and it should be addressed. There are many algorithms that can detect forgery, thorough the study, the most reliable approach or algorithm for detecting forgery is Convolutional Neural Network (CNN), but the singular approach of CNN is not much reliable. We hereby present a deep learning hybrid approach of Error Level Analysis with convolutional neural network for detecting Image manipulation. To detect whether an image is manipulated, the ELA from the image is extracted and will send the information to CNN. The training of the above algorithm will be performed on the dataset CASIA V2.0 which contains many forged images as well as authentic image. Transfer learning with UNET (encoder-decoder model) has also been applied for image segmentation. The experimental result on the dataset used are quite good, and these methodologies have been duly evaluated. The outcome of the proposed model outperforms the other well-known Transfer learning approaches compare to the findings. The study suggests that proposed model can get up to 0.910 and 0.930 accuracy and precision, respectively. The comprehensive analysis comparatively presented in this study provides valuable information that can guide the development of precise forgery detection technologies, safeguarding the honor of digital content and lessening the consequences of picture frame-ups in the rapidly evolving world of digital media.