Effective Steganalytic Algorithm based on Residual Value of Pixels in an Image

Steganalysis is the art and science of detecting the presence of secret information in the cover object. In this work, we propose a steganalytic algorithm, which differentiates the normal image from the stego image. To achieve this, we train a set of images by applying a model that finds all the dependencies among pixels. This model, then calculates the residual value along 0, 45, 90 and 135 degrees. The calculated residual value is then converted into a single digit for memory reduction. These residual values are considered as features. In testing phase, the residual value is calculated for all the pixels and the training feature value is loaded. SVM classifier is employed to differentiate between the training and testing feature. On the basis of minimal distance value, it is decided whether or not the image is stego.