A comprehensive review of deepfake generation methods, detection strategies and open challenges in multimedia and Internet of Things systems
TL;DR
This work surveys deepfake generation and detection methods, and highlights the use of DL and neural networks in comparing detection strategies across computing and IoT systems.
Abstract
Deep Learning (DL) has driven extensive progress in Artificial Intelligence (AI) and data analysis in areas such as communication and computer vision but this progress also increased misuse. A significant example is deepfake technology, which produces highly realistic fake videos, images, and audio. Such content is deceptive and risks privacy, trust, and national security. Deepfake attacks are now common in multimedia processing and are spreading to Internet of Things (IoT) edge devices. As IoT systems handle large amounts of on-demand multimedia traffic in innovative applications, they are vulnerable to deepfake threats. Hence, automatic detection is essential to ensure authenticity in communication. This work surveys deepfake generation and detection methods. The survey focuses on recent studies. It examines research directions and applications in IoT systems. The survey draws on publications from 2018 to 2025. It highlights the use of DL and neural networks in comparing detection strategies across computing and IoT systems.