The Effect of Correlogram Properties on Blind Steganalysis in JPEG Images

Abstract

Blind image steganalysis is a technique for discovering the message hidden in images in an independent manner than embedding the hidden message. The content of the image contribute to the success of steganalysis drastically. In the past, texture, one of the most basic features in any image processing, has been used for content-based image classification. Correlogram properties as textural based features have numerous applications in this field. Homogeneity, contrast, correlation, energy and entropy are the correlogram properties used more frequently than others for this purpose. In this article, the impacts of these properties, as descriptors for image content, on blind steganalysis in JPEG image are investigated. The results indicate that when correlogram homogeneity increases, the false image detection of blind steganalysis increases accordingly; while, decrease in correlogram contrast and entropy, leads to an increase in error. The energy and correlation of correlogram have unspecified effects on image blind steganalysis.