Histogram estimation-scheme-based steganalysis defeating the steganography using pixel-value differencing and modulus function

Abstract

We propose a steganalysis defeating the steganographic method using pixel-value differencing and modulus function, which is a recent method with high security and capacity for secret communication. The presented steganalysis is designed to reveal the existence of the message and uses three steganalytic measures that remarkably increase their values in the stego images. Hence, the stego images are statistically separated with the cover images. Detection of the hidden message is possible by modeling the changes generated by the embedding process and comparing the values of the steganalytic measures. To increase the performance of the steganalytic measures, a novel histogram estimation scheme is used to estimate the histogram value of the cover image and the embedding ratio. A support vector machine classifier is adopted to discriminate between cover and stego images. The experimental results verify that the proposed steganalysis can detect the stego images with 97.1% accuracy, even though the embedding ratio is just 10% of the maximum hiding capacity. Also, the length of the hidden message can be successfully estimated without the cover image.