A Comprehensive Review of Keypoint-Based Copy-Move Forgery Detection: Challenges and Advances

Copy-move forgery detection (CMFD) remains a significant issue in digital picture forensics, as it can conceal altered areas by duplicating and modifying them. This paper provides a comprehensive evaluation of keypoint-based CMFD techniques, meticulously categorizing the literature into four main groups: classical methods, efficient and lightweight detectors, hybrid approaches, and deep learning–enhanced models. To provide a more comprehensive picture, we also discuss new trends and make comparisons, focusing on recurring issues such as keypoint sparsity, high computational cost, and dataset bias. We also discuss promising areas, such as transformer-based frameworks, adversarial robustness, and lightweight self-supervised learning. Additionally, a list of regularly used datasets and assessment metrics is provided to facilitate studies that can be replicated and enable fair comparisons. This study provides researchers with a structured reference by compiling existing advances and thoroughly evaluating their merits and drawbacks. It shows how to make CMFD systems in the real world more durable, scalable, and valuable.