Identifying Botnets by Analysing Twitter Traffic during the Super Bowl

Salah Safi, Huthaifa Jawazneh, Antonio Mora, Pablo García, Hossam Faris, Pedro Castillo


Detecting accounts broadcasting illegal contents at sporting events in social networks is an important problem of difficult solution, since the traditional intrusion detection systems are not effective in online social networks due to the speed with which these kind of messages and contents spread. Thus, there is an increasing need for an adequate and efficient detection system of the so-called botnets used for the distribution of illegal contents on online social networks. In this paper we propose using well-known classification methods to analyse the activity of Twitter accounts in order to identify botnets. We have analysed the Twitter conversations that include hashtags related to the Super Bowl LIII (February 3, 2019). The objective is to identify the behaviour of various types of profiles with automatic and non-standard spamming activities. In order to do so, a dataset from public data available on Twitter that includes content published by human-managed accounts and also by bots that used hashtags related to the event has been collected. This dataset has been manually labelled to create a training set with tweets posted by humans and bots active in Twitter. As a result, several types of profiles with non standard activities have been identified. Also, some groups of accounts have been identified as botnets that were activated during the Super Bowl LIII (2019).


Paper Citation