Detecting the unmasking features of a social bot at the syntax level of a generated message

Abstract

The article aims at identifying the unmasking signs of bots based on the results of the analysis of the syntax of messages generated by them. The matter concerned is of high relevance due to the following reasons. First, the results of the annual analysis of Internet traffic indicate that there is no decrease in the traffic-generated by malicious bots. Second, the activity of highly organized bad bots poses a high level of informational threat to the psychological safety of citizens. Altogether, these factors, including the high level of anxiety among citizens, which has been noted by sociologists over the past few months, transfer the threat to the information and psychological security of citizens into the category of challenges for the digital sovereignty of the state. Methods for detecting social bots, presented in previous works, involve the collection and processing of a large amount of data, including metadata of the Internet QMS user profile, data on the time and period of publication of messages, analysis data of the Internet traffic of the Internet QMS user, e.t.c. Despite their high performance indicators, these methods have a common drawback. Big data collection and processing consume a significant amount of time. Several unmasking signs of a bot identified in this work can be used as the basis for a qualitatively new method for detecting social bots. The method is based on the linguistic characteristics of the messages generated by social bots. A distinctive feature of the method will be the relatively small amount of information needed to identify a social bot based on the results of a quantitative analysis.