Hybrid Deep CapNet-VGG19 Model for Detecting Forged Images and Videos

In recent days, Video and image forgery detection becomes significant aspect due to modern editing software which offers powerful and easy to handle tools for the manipulation of videos. a technique of video detection faces much more challenges like video dataset, post-processing issues, noise, computational time/complexity, deepfake detection approach and insufficient anti-forensic. Consequently, this concern is addressed in this work by presenting a deep capsule network model based forensic detection. For this purpose, this paper presented a Deep learning (DL) based model for the detection of forged images and videos. This model presents a technique which uses a Deep capsule network (CapsNet) and VGG19 model for detecting forged videos and images in a huge range of forgery scenarios, which includes detection of replay attack and computer-generated video/image detection (both fully and partially). The performance outcome is estimated for deepfake dataset and outcomes attained are compared with existing models in terms of accuracy at both frame level and video level. The analysis shows that proposed model is effective and offers enhanced outcome than traditional models.