<p>Image tampering detection (ITD) represents a critical area of research that addresses the challenges posed by advanced digital alteration tools in evaluating the authenticity and integrity of visual media. This review conducts a thorough analysis of ITD techniques, categorizing them into two primary types: Active methods, which utilize embedded information for verification, and Passive methods, which analyze image characteristics without prior embedding. It highlights the transformative impact of deep learning architectures such as convolutional neural networks (CNNs), encoder-decoder models, and recurrent neural networks (RNNs) in enhancing both tampering localization and classification accuracy. Furthermore, this analysis examines the reliability of recent innovations, including hybrid models and feature extraction techniques, especially when applied to benchmark datasets such as CASIA and CoMoFoD. These datasets facilitate rigorous evaluation based on metrics such as precision, recall, and F1-score. Despite significant progress, challenges persist in developing algorithms that are resilient to sophisticated manipulation techniques, achieving real-time processing efficiencies, and enabling lightweight implementations suitable for edge devices. The review underscores the importance of interdisciplinary collaboration and the integration of explainable artificial intelligence in advancing digital image forensics across various applications, including journalism, forensic investigations, and ensuring the integrity of social media content.</p>