Text Classification of Illegal Activities on Onion Sites

Abstract

Onion sites work using the Hidden Service Protocol, which helps to keep a double anonymity. A such system allows sites to place malicious and illegal content. An identification and tracking of such resources is an important problem, that’s why the article sets a task of developing a system for accurate thematic classification of textual content blocks of hidden web pages using k nearest neighbors method. The article presents the method of content separation placed on Russian-language onion-sites. The research illustrates the analysis of text categorization results based on collected dataset for the implementation of machine learning.