Detection of Tampered Real Time Videos Using Deep Neural Networks

Abstract The creation and sharing of videos with the aim of promoting the use of digitally interactive multimedia, such as music, graphics, and video between devices for both social networking applications and everyday tasks have increased significantly along with the use of digital communication devices in recent years. In the digital sphere, forgery techniques and motivations have substantially evolved. Before, methods of video editing were applied to improve digital data. With the rise in popularity of low-cost, user-friendly video editing software, there are a number of downsides and risks related to these editing methods. In order to produce altered or fraudulent videos, additional footage is mixed, edited, or synthesized. Existing method uses methods that detect forgery in videos with simply static backgrounds only. Proposed systems uses a deep learning strategy that incorporates transfer learning utilizing VGG16 and Customized CNN layers to categorize real time videos as tampered or authentic .With the aid of deep neural networks, the suggested method may identify forgery in films with both static and moving backgrounds. The experimental findings show that the suggested strategy is over 99.9% more accurate and effective than existing methods also it provides trustworthy results with low computing cost and strong detection performance.