Social Network Users Profiling Using Machine Learning for Information Security Tasks

Abstract

The need for bot detection is growing in proportion to the increase in the number of social network users. The robotization of processes has not escaped social networks, with the result that bots, designed to mimic human behavior, create a burden and, in some cases, threats to users, including manipulation and misinformation. Classical information security threats related to bot activity are DDoS, collection and distribution of user data, manipulation of billing systems, and misuse of services. Often bot technology is used for scoring bonus points or using other customer loyalty mechanisms to gain their own benefit, in violation of the service policy. The problem is that it is often hard to confirm the correspondence between a real person and a profile due to the large amount of disparate information about users' activity, as well as the use of modern technologies, including machine learning, to develop bots. This paper focuses on the problem of detecting bots in social networks using machine learning. We propose an automatic, retrainable method for detecting fake accounts on a social network. The study describes the result of developing user classification models based on the activity logs of social network users in the problem of automated user profiling, that is, determining whether a user account is genuine or a bot is hiding behind it. The aim of the work is to develop methods for detecting bots using machine learning and intelligent analysis. In our work to solve the problem we use gradient boosting with an accuracy of AUC = 0.9999.