Disjunctive Edge Map based Image Sterilization for Destruction of Steganograms in Spatial Domain

Image sterilization is a form of steganalysis where the suspected carrier image is actively manipulated, filtered or modified to mutilate the steganograms likely to be concealed in it. This paper proposes an image sterilization method for mutilating the secret messages hidden in a cover image by adaptive pixel modification on the basis of an edge map. The edge map is obtained from the disjunctive combination of binary thresholded edge matrices of Canny, Weighted Sobel and Laplacian filtering on the suspected stego image. The sterilization scheme injects high impact multiplicative noise to the edge pixels while the non-edge pixels are modified using low impact additive noise. Eight benchmark steganography algorithms including LSB, PVD, EMD, WOW, HUGO and UNIWARD are applied to sample images taken from USC-SIPI database to prepare the experimental dataset. Analysis of experimental results reveals the superiority of the proposed technique in comparison with the state-of-the-art image sterilization methods.