Multiple forgery detection in digital video with VGG-16-based deep neural network and KPCA

Abstract

The amount of video data is growing exponentially on a daily basis. Easily available soft ware or mobile applications offer simple tools to perform the forgery in the video. So, before sending these videos from one place to another, it is important to verify them. In this paper, a forgery detection system is proposed to detect the multiple forgeries in the video using the VGG-16 deep neural model and KPCA (Kernel Principal Component Analysis). The proposed system works in four stages. The preprocessing approach is initially employed to extract and resize video frames. Then, a pre-trained VGG-16 model is tuned to extract the visual features from each input frame. A feature selection methodology, such as KPCA, is applied to minimize the dimensions of extracted features. Finally, correlations distribu tion among the selected features is analyzed to expose the forgeries. The performance of the proposed system is tested on a forged video dataset. The simulation result reveals that it gives better performance in identifying forgeries in the video, with accuracy and pre cision of 97.24% and 96.86%, respectively. In addition, the significance of the proposed system is that it yields superior results in post-processing operations like noise addition and adjustments to brightness, contrast, and hue.