The increasing rate of development of deepfakes, advanced editing software, and generative AI models has made it challenging to distinguish between real visual content and forged or artificially created content. The forged images and deepfakes tend to be very realistic, and thus manual verification and traditional forensic analysis are rendered ineffective. This results in the spread of misinformation, online fraud, and a lack of digital trust, thus making the need for an automated and intelligent forgery detection system imperative. This work introduces IVFD: An Intelligent Visual Forgery Detector for Artificial Visual Content Identification, a deep learning-based forgery detection system capable of detecting forged or artificially created images and videos. IVFD employs convolutional neural networks (CNNs) and feature extraction techniques to detect visual anomalies in images and videos, including pixel errors, texture, lighting, motion, and boundary anomalies. Images and videos are fed into a carefully designed preprocessing and classification pipeline, which examines critical features and produces an authenticity score reflecting the likelihood of forgery. The proposed system was tested using benchmark forgery datasets for images and videos, and various CNN models were compared to determine the most effective model. The experimental results show that IVFD has a high accuracy rate in image forgery and deepfake video detection. In conclusion, IVFD is a fast, accurate, and automated solution for digital forensics, media verification, and online content authentication. Future improvements can be considered in the form of transformer models, autoencoders, and video forgery analysis using LSTM.
IJERST (Fri,) studied this question.