Detecting Passive Image Forgery

• Tech Stack: Tensorflow, Keras, Python, Opencv
• Research Paper: Link
• Github URL: Project Link

Today, rapid advances in science and technology have made it possible to manipulate multimedia content easily with a variety of editing tools. The integrity and trustworthiness of multimedia content are seriously threatened by this. As a result, sustaining digital images is increasingly important since they contain important information and are used as proof in a variety of sectors. A number of researchers have developed different image forensics detection methodologies due to the requirement and importance of digital image forensics. The basis of image forensics is passive image forgery detection. Passive forgeries such as image splicing, copy-move, and retouching are often used techniques that compromise the authenticity of an image. Recently, a huge amount of study has gone into creating innovative techniques for identifying various image forgeries. In this paper, we present a new methodology for detecting image forgery that accepts an RGB image as input and converts it to an ELA (error level analysis) image before classifying whether the image is tempered or not. This model is a mix of modified DenseNet169 and modified DenseNet201.Through rigorous testing and validation, our model demonstrates a superior capacity to identify tampering even when there are realistic and subtle changes created by sophisticated image editing tools. In our investigation, we achieved an accuracy of 93.79 on test images.