Adaptive Text Steganography by Exploring Statistical and Linguistical Distortion

Most state-of-the-art text steganographic techniques keep the semantic of cover texts almost unchanged based on synonym substitution, but some statistical features of cover texts will be utilized by steganalyzers to detect the existence of the stego texts. In this paper, an adaptive text steganography is proposed to minimizing the cost caused by synonym substitution, in which the cost function is defined as a hybrid of statistical distortion and linguistical distortion (HSL). Furthermore a double-layered STC (Syndrome-trellis code) scheme is designed to minimize distortion of synonym substitution. Experiments show that HSL can achieve higher secure level compared with traditional synonym substitution based methods.