A Method for Quantitative Steganalysis Based on Deep Learning
Yu Sun, Tianyun Li · 2019
In this paper, a quantitative method of image steganalysis for LSB matching is proposed. It applies Ye network (YeNet) which has the best performance in field of binary steganalysis at present. According to experimental observation that through YeNet the final feature output of stego image changes regularly with the embedding rate, the mapping relationship between embedding rate and feature distribution can be fitted, so as to estimate the embedding rate of a given stego. Compared with previous researches, this method has a better performance and the advantage is more pronounced for low embedding rates.