Detection of identity swapping attacks in low-resolution image settings

While significant advances have been made in high-resolution and well-controlled face images/videos, low-resolution face analytics is a much more complicated and yet unsolved problem. On top of that, if the face images occurring in the low-resolution videos are fake (especially deepfake), then detecting the authenticity of those faces becomes exceptionally challenging. In the literature, several works have been proposed for deepfake detection on high-resolution images. However, no studies tackle the vital aspect of low resolution. In this research, we address this issue and propose a first-ever low-resolution identity swap attack detection algorithm. We assert that due to less information content, even a complex architecture might not be able to learn an effective decision space. Therefore, a novel artifacts amplification and classification algorithm is proposed to handle the lack of information content. We report our results using extensive evaluations using multiple databases, resolution settings ranging from very low-resolution face images of size ($16\times 16$) to medium resolution ($128\times 128$), and attack types. These extensive experiments demonstrate the strength of the proposed algorithm and its effectiveness in making it ready for in-the-wild settings. Our results show the novel findings and the superiority of the proposed algorithm compared to existing state-of-the-art works.