Steganalysis of skin tone images using textural features

Abstract

Around 700 million selfies are posted daily on social media. Skin tone based steganography refers to a steganography method where the confidential message data is incorporated within the skin tone region of images. Skin tone area of an image provides an excellent place for data hiding. The objective of this work is to detect whether any message is hidden in the skin portion of image. Different methods like LSB based, signature based etc. exists for steganalysis of images that concentrates on the entire image. This work focusses on a steganalytic technique based on textural features of skin tone images. The statistics on the texture of human skin acquired from cover and stego images are used for creating a trained classifier model and then tested using three classifiers. Existing tools like StegHide, Outguess etc. use different methods for hiding information in images. Various steganalytic techniques exist to detect messages concealed by means of above tools, which give an accuracy range between 85% and 96%. As the complexity of hiding is increased by embedding in transform domain, the existing detection rate using GLCM is 91.79%. From the experimental results, it is observed that an accuracy of 93% is obtained and the proposed technique outperforms existing methods in terms of detection rates.