Multi-scale Extracting and Second-Order Statistics for Lightweight Steganalysis

Steganography is a technology that modifies complex regions of digital images to embed secret messages for the purpose of covert communication, while steganalysis is to detect whether secret messages are hidden in a digital image or not. In recent years, it has become necessary to deploy computer vision based algorithms on devices that are mobile or have limited computational memories. However, the emergence of steganalysis prove the point that the more parameters available to the model, the better the presentation will be. In order to enable the model to achieve the extraction of steganographic noise with a tiny number of parameters, this paper proposes a lightweight steganalysis algorithm based on multi-scale feature extraction and the fusion of multi-order statistical properties. Compared with Yedroudj-Net, Zhu-Net and SRNet with S-UNIWARD embedding rate of 0.4 bpp, the numbers of parameters was decreased by 87.9%, 98.2%, 98.9% correspondingly and the accuracy of steganalysis was improved by 7.12%, 2.65%, 3.42% respectively. Our experimental results show that the model not only has a reduced number of parameters for existing steganalysis, but also can effectively boost the accuracy of steganalysis.