Video Tampering Detection in Real Time

Videos are used everywhere nowadays for various purposes. There is great improvement in visual innovations, which has wide applications. Videos are essential in almost all industries such as surveillance cameras, social network, e-learning, advertisements, medical etc. Though the rise of digital technology made our lives easier, tampering the videos is done with ease nowadays. These changes reflect in our lives and may affect some decisions as digital videos play a vital role in every aspect. Thus we need to consider the authenticity of the video. Video tampering detection is a method to detect whether the video contents are original or modified. In this paper, to detect the forged content in the videos, we propose a deep learning approach. The aim of the model is to detect whether the video is tampered or not. The proposed method classifies the video as authentic or tampered by performing patch analysis and error level analysis using video frames and using a deep learning classifier. Patch model includes preprocessing the video frames and then separating the patches whereas in error level analysis, the video frames undergo compression level at 90% and then difference is applied. The deep learning classifier Resnet50 is used for both the methods. By applying these two techniques, video forgery can be detected.