Social Media Origin Based Image Tracing Using Deep CNN
Nadia Siddiqui, Areesha Anjum, Mahreen Saleem, Saiful Islam · 2019
determining information regarding provenance of the digital images is challenging for image forensic scientific community as it employs tracing of the crucial yet hidden signatures associated with the images. Improving image content's history can be deliberated to mark an investigation for the assessment of the image by knowing its acquisition device or, model as the basis of investigations in certain direction. The fundamental approach for going back to image's provenance would involve revealing every subsequent postprocessing, like filtering or an interpolation achieved, applied on the uploaded image. This paper presents a model to uncover such distinct traces by exploiting Deep Learning (DL) based methods to examine the image origin in social networks, and particularly attempts to perceive from which social network the specific image has been downloaded. Here DL- based effective Convolutional Neural Networks (CNNs) is proposed to differentiate the image path in corresponding social media platform. In this work images uploaded/ downloaded from multiple social media paths are investigated to predict the image with correct class of media path. Along with source path identification, this work is expanded to carry out the social media distinction in spite of diverse JPEG quality factor.