A Robust Copy-Move Forgery Detection In Digital Image Forensics Using SURF

In recent years, due to the technological revolution in editing digital images, various advanced image manipulating software has been used to build new unrealistic images without leaving traces of what happens, therefore tampering will be hard to detect visually. Digital image forgeries have many forms but still recognizing copy-move forgery is very challenging. Hence, this paper introduces a new robust algorithm to detect copy-move forgery based on Speeded Up Robust Feature (SURF) descriptor, Approximate Nearest Neighbor (ANN) as a feature matching, Simple Linear Iterative Clustering (SLIC) used as a clustering algorithm to divide the whole image into superpixel blocks. The doubted regions are determined by replacing the matched feature points with corresponding superpixel blocks then the neighboring blocks have been merged based on similar Local Color Features (LCF). Finally, morphological close operation applied to elicit the doubted forged regions. Proposed algorithm recorded a running time of 3.84 seconds with 91.95% localization accuracy applied on various datasets such as CoMoFoD, MICC-F2000, MICC-F220, and MICC-F600 for detecting tampered plain copy-move, duplicate regions, post-processing and pre-processing attacks like color reduction, blurring, brightness modifications, noise addition, geometric attacks, and JPEG compression as an evaluation of robustness.