This is the authors' abstract. We don't add key points for this paper.
The detection of manipulation of multimedia data can be categorized as active and passive approaches.1 The active approaches use source information, such as watermarks and digital signatures. For instance, Bahirat et al.2 proposed a watermarking-based framework for authentication and localization of tampering in red, green, blue (RGB) and the 3-D point cloud. The watermarking methods, which are computationally expensive, cannot be applied on raw unprocessed data. Passive approaches, or blind forensics,3 are intended for testing multimedia data where the original or source information is not available. Deep-learning-based tampering detection avoids the need to perform various forensic tests to detect whether multimedia data have been manipulated or not.4 However, deep-learning-based methodologies need to balance false positives and negatives, as pointed out by Bayar and Stamm.4 Apart from the need for detecting manipulations in multimedia data, there is also a need for localizing (i.e., identifying the region) where the manipulation occurred.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Balakrishnan Prabhakaran (2024) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: