Video Origin Camera Identification using Ensemble CNNs of Positional Patches

The use of multimedia devices has increased immensely with the availability of affordable mobile loT technologies. Consequently, video capturing applications are incorporated in many crucial sectors including crime investigations and surveillance. In such applications, it is important to ensure the integrity of the video data and verify its source for digital forensics purposes. In this paper, we tackle the problem of video source camera identification to validate the origin of video data that may come from numerous types of cameras. Specifically, we propose a novel approach based on ensemble of convolutional neural networks (CNNs) for detection of video source. In our approach, the video is analyzed using patches obtained from I-frames after splitting each frame into quadrants and training a CNN per quadrant for location awareness. Our experimental results demonstrate that our model brings significant improvement for video source identification accuracy compared to existing machine learning based techniques.