Video Forgery Detection using CNN

With the widespread use of digitally interactive multimedia such as audio, images, and video, there has been a significant increase in the mode and motivation to create digital forgeries. The widespread availability of video information and services, as well as the low cost of devices such as cameras, camcorders, and CCTVs, has led to widespread use of video information and services in our society for a variety of purposes such as video surveillance, forensics investigation, and entertainment. Previously, video editing techniques were mostly employed to improve digital information. However, as the popularity of low-cost, easy-to-use video editing software has grown, so has the number of negative repercussions and risks associated with such editing procedures. By merging, changing, or synthesising new footage, video forgery is a technique for creating changed or fraudulent videos. A method based on deep learning is given in the proposed system for classifying videos as tampered or original. The video clip that is used as input is divided into two categories: original and modified. The video is segmented into non-overlapping frames, and the authenticity of the movie is determined by whether or not all of the frames are genuine. The suggested method uses a deep CNN model that has two types of layers: (1) CNN layers which involve convolutional, pooling and fully connected layers and (2) Parasitic layers.