Recent Advances in Digital Image and Video Forensics, Anti-forensics and Counter Anti-forensics

TL;DR

The paper tackles the rise of manipulated multimedia by providing a comprehensive survey of image and video forensics, including the growing area of generative media and anti-forensics. It covers source identification and forgery detection for both camera-generated and GAN-generated content, across still images and video, and extends to anti-forensics and counter anti-forensics techniques. The contributions include an up-to-date synthesis of methods, datasets, and open problems, with emphasis on the challenges posed by GANs and the need for robust, generalizable detectors. The work highlights practical implications for authentication in media-rich contexts and underscores the importance of legal admissibility and explainability in forensic conclusions.

Abstract

Image and video forensics have recently gained increasing attention due to the proliferation of manipulated images and videos, especially on social media platforms, such as Twitter and Instagram, which spread disinformation and fake news. This survey explores image and video identification and forgery detection covering both manipulated digital media and generative media. However, media forgery detection techniques are susceptible to anti-forensics; on the other hand, such anti-forensics techniques can themselves be detected. We therefore further cover both anti-forensics and counter anti-forensics techniques in image and video. Finally, we conclude this survey by highlighting some open problems in this domain.

Figures (2)

• Figure 1: GAN Generated Images from Wang_2019 based on StyleGAN works by Karras_2019Xuan_2019.
• Figure 2: Organization of the image and video forensics