A novel text steganography by context-based equivalent substitution

The text steganography is a prevalent technique for covert communication. We propose a novel linguistic steganography by context-based substitution. We utilize the conditional equivalence between the definite article and demonstrative adjectives ubiquitous in natural texts to embed covert messages. Considering the diverse usage of these words, we adopt a part-of-speech filter, a 3-gram language model, a 2-range context, the minimum matching distance, and the perplexity to evaluate the substitution suitability degree in our steganography and the substitution can only be carried out under certain conditions. The results of the blind test show that the generated stego-text can hardly have syntax errors or unsuitable words, which proves that the steganography is imperceptible.