Аннотация
This chapter presents a comprehensive examination of deep learning approaches in copy-move forgery detection (CMFD), a critical domain in digital image forensics. As image manipulation techniques become increasingly sophisticated, the need for robust and adaptive detection methods has never been more pressing. We analyze state-of-the-art deep learning architectures, including convolutional neural networks, autoencoders, and generative adversarial networks, elucidating their principles, strengths, and limitations in CMFD tasks. Our investigation reveals significant advancements in detection accuracy, robustness to geometric transformations, and the ability to handle multiple forgeries within a single image. We provide a systematic comparison of leading methods across various datasets and evaluation metrics, offering insights into their relative performance and applicability. The chapter also addresses critical challenges in the field, such as the need for standardized benchmarks, computational efficiency, and model interpretability. By synthesizing current research and identifying promising directions for future work, including the integration of explainable AI and the development of quality-independent methods, this chapter serves as a vital resource for researchers and practitioners in digital forensics, computer vision, and cybersecurity.