A statistical attack on a kind of word-shift text-steganography

Text-steganography plays an increasingly significant role in covert-communication on Internet. Compared with study on text-steganography, research on text-steganalysis is in its infancy. In this paper, we present a statistical analysis of a kind of word-shift text-steganography by using neighbor difference (length difference of two consecutive spaces). We classify nature PDF documents into two cases serving for our analysis and statistically model wordshift steganographic embedding situations. Our method can reliably detect existence of hidden information. Furthermore, it quantitatively gives estimated embedding rate. The high performance of our method is demonstrated by experiments. Our work will contribute to both text-steganalysis and text-steganography.