Blind detection of image splicing based on run length matrix combined properties

Image splicing is a technique commonly used in image tampering. In order to achieve image splicing blind detection, a blind, passive, yet effective splicing detection method is proposed in this paper. In this method run length matrix is used to extract image feature and generate the identification model with combination of Neighborhood DCT Coefficient Co-occurrence Matrix Feature and Markov Feature. Support vector machines (SVM) also is selected as classifier for training and testing while genetic algorithm is used to optimize parameters based on evaluation criteria AUC. Experimental results show that there is high classification accuracy for obtained model by this method.