Abstract

The rapid proliferation of manipulated visual content has raised critical concerns about authenticity, privacy, and public trust in digital media. Traditional forgery techniques such as image splicing and copy–move have long posed challenges for forensic analysis, but the advent of AI-driven generative models has introduced a new era of sophisticated and realistic forgeries, including deepfakes and image morphing. In parallel, the continuous evolution of deep learning has provided powerful tools to counter these threats, making it a cornerstone in modern image forgery detection. This review focuses on the application of deep learning methods to detect image forgeries over the past eight years. The paper covers a wide range of approaches addressing both traditional and modern forgery types, discussing representative models, methodologies, datasets, performance results, and their respective contributions and limitations. The goal is to provide a clear understanding of current progress and to highlight key challenges and opportunities for future research in deep learning-based image forgery detection.