The level of accuracy of author attribution of a text is not high enough at the lexical and syntactic levels of a language as these levels are not strictly organized systems. In this study, the author attribution of a text is based on the differentiation of phonostatistical structures of styles. We have developed a system of differentiation of phonostatistical structures of styles, which differs from the existing ones by the chosen level of a language ‒ phonological. At this level of a language one can obtain results with a greater accuracy. In addition, the system constructed is based on a modular principle, which makes it possible to rapidly modify the developed software. We have developed methods and models that are based on the theory of mathematical statistics and allow the improvement in the accuracy of differentiation of phonostatistical structures of styles. A method was devised for a comprehensive analysis of phonostatistical structures of styles, as well as a multifactor method for determining the degrees of action of factors related to style, substyle, and author's manner of presentation. We have constructed a statistical model of stylistic differentiation using the ranking method, and a statistical model for determining a general stylistic markedness of the examined text. A software system for the differentiation of texts was designed. The criterion for the differentiation of texts is the mean frequencies of groups of consonant phonemes. In the process of implementing a system we used the programming language java, which ensures that the software is platform-independent. This study reports results of the application of the developed methods, models, and software tools. The research results confirm that author attribution of a text at the phonological level is more effective. The developed methods, models, and means for the author attribution of a text could be used when determining the percentage of creative contribution of each of the co-authors of scientific papers.