On the Characterization of Digital Trolls from Twitter Big Data

Recently, Twitter becomes one of the most common and effective social media tools in our life. People use Twitter to share their opinions, feeling, and orientations, especially during political unrest or protests. In order to disrupt the protest operations on Twitter or to influence public opinion, electronic flies (Trolls) are widely and effectively used.Accordingly, there is a need to find a method that can automatically and precisely detect these accounts and isolate them from Twitter. Moreover, detecting and characterizing these accounts becomes a significant task to reduce or mitigate its effect on the real general opinion. Thispaper presents an intensive analysis that can be utilized to effectively detect the troll accounts and isolate its bad effect from Twitter.We considered the public trolls accounts datasetspublished by Twitter and we also gathered a new dataset from Twitter that includes tweetsand users’ information from different countries to make a fair analysis for the trolls’ accounts.  The results show thatthe suspicious activities of Twitter troll accounts can be used to detect most of these accounts automatically without using sentiment analysis and opinion mining techniques with accuracy of 95%.To accomplish this task, we propose a set of robust and efficient features that can accurately characterize troll accounts with a relatively small number of features.