Improving N-Gram linguistic steganography based on templates

The automatic generation of stegotexts is an area of research within linguistic steganography with a great interest. Different proposals have been published about the development of stegotexts with statistical and linguistic validity in natural language but, unfortunately, the stegotexts generated following some of these procedures suffer from different types of lexical, syntactic, semantic problems or problems of global coherence that would eventually lead the automatically generated texts to be detected by a human reader. This article presents an improved implementation of an algorithm of N-Gram statistic imitation and introduces a new concept of manual edition of the stegotexts generated, based on the idea of templates. This procedure permits the creation of high quality stegotexts to hide a small quantity of information, hundreds of bits, useful to distribute a symmetrical key, a short message, an url, etc. The procedure developed in the article can be applied to different languages, but here, its usefulness is demonstrated for Spanish. The Stelin tool is presented free at http://stelin.sourceforge.net.