A systematic literature review on image splicing detection and localization using emerging technologies

Computer vision applications involving digital forensic investigations widely use digital images as legal documentary proof. Ensuring the authenticity and reliability of digital images by locating potential tampering is a critical area of concern for forensic applications. Splicing is one of the most commonly used methods for image tampering in digital domains. This systematic literature review (SLR) is conducted using PRISMA guidelines to explore the opportunities and challenges in the image splicing detection and localization (ISDL) domain. A total of 99 empirical papers were selected for an in-depth review from four major databases: IEEE Explore, Science Direct, Springer, and Web of Science. Papers were selected based on specific inclusion and exclusion criteria, focusing solely on those addressing deep learning, machine learning, transfer learning, and quantum computing technologies. The survey was conducted by framing 8 research questions based on ISDL to identify potential answers to its implementation. The synthesis shows that 83.84% of the ISDL studies were based on generic applications, and 73.74% of the studies utilized machine learning models for classification. Furthermore, accuracy, F1-score, and sensitivity were the most preferred evaluation metrics used by 63.64%, 38.38%, and 36.36% of studies, respectively. For ISDL applications, 54.54% of the studies used CASIA TIDE v2.0, and 37.37% incorporated graphics processing units (GPU). This paper presents a thorough synthesis of existing studies in the ISDL domain while highlighting the importance of various emerging technologies in dealing with digital forensics. The outcomes are expected to be useful to industry experts, researchers, and policymakers to establish safe practices in computer vision applications.