Identification of source social network of digital images using deep neural network

Abstract

Listen

Translate article

• A novel method is presented for the identification of image provenance. • It identifies if the image is taken with a camera or it has been downloaded from a specific social network. • It does not rely on hand-crafted fingerprints to learn the social network specific traces. • Transfer learning approach is used to learn the traces from the images transformed to the DCT or DWT domain. • The proposed method is robust in capturing the traces from the resized images. Identification of image provenance is one of the crucial tasks to be tackled in the field of image forensics . Due to the rapid growth in technology, a huge amount of digital images are shared among users through social networks which facilitates many malicious activities . Establishing the provenance of downloaded images is an important task that could help focus investigations in a specific direction. Such a task is based on the identification of unique fingerprints that are imprinted on images by the social network during the process of upload and download. Based on this, we propose a deep learning based approach that learns the unique traces from the images transformed to the discrete cosine and wavelet domains , to investigate whether the image under test originates directly from a camera or from a specific social network site . The proposed method is able to efficiently identify the specific social network of provenance of the downloaded images and outperforms the state-of-the-art techniques. The encouraging results on images represented in the wavelet domain obtained for the well-known datasets, namely, VISION, IPLAB, and FODB sheds the light into a new way of identifying the image provenance.