A SURVEY OF DEEPFAKE IMAGES DETECTION MODELS
TL;DR
This comprehensive review underscores ongoing efforts to combat image manipulation and safeguard the authenticity of visual media, and highlights the effectiveness of deep learning-based approaches, which excel due to their ability to process large and complex datasets, perform end-to-end learning, adapt to new challenges, and offer robustness, scalability, and continuous improvements.
Abstract
The rapid progress in artificial intelligence (AI), machine learning, and deep learning has led to the development of innovative tools for multimedia content manipulation. While these technologies have legitimate applications in entertainment and education, they have also been misused to create Deep fakes highly realistic fake videos, images, and audio. Deep fakes are often exploited to spread misinformation, propaganda, and political discord, as well as for harassment and blackmail. In response, researchers have developed various detection techniques, including machine learning, spectral analysis, steganography, and feature extraction. This paper presents a systematic literature review of 56 relevant studies published between 2007 and early 2025, analyzing different forgery detection methods. The review highlights the effectiveness of deep learning-based approaches, which excel due to their ability to process large and complex datasets, perform end-to-end learning, adapt to new challenges, and offer robustness, scalability, and continuous improvements. Additionally, widely used benchmark datasets such as Face Forensics++ and Celeb-DF have played a crucial role in advancing research in this domain. Standardized evaluation metrics ensure consistency in model comparisons and facilitate systematic progress tracking. This comprehensive review underscores ongoing efforts to combat image manipulation and safeguard the authenticity of visual media.