Still out there: Modeling and Identifying Russian Troll Accounts on Twitter

Abstract

There is evidence that Russia's Internet Research Agency attempted to interfere with the 2016 U.S. election by running fake accounts on Twitter-often referred to as "Russian trolls". In this work, we: 1) develop machine learning models that predict whether a Twitter account is a Russian troll within a set of 170K control accounts; and, 2) demonstrate that it is possible to use this model to find active accounts on Twitter still likely acting on behalf of the Russian state. Using both behavioral and linguistic features, we show that it is possible to distinguish between a troll and a non-troll with a precision of 78.5% and an AUC of 98.9%, under cross-validation. Applying the model to out-of-sample accounts still active today, we find that up to 2.6% of top journalists' mentions are occupied by Russian trolls. These findings imply that the Russian trolls are very likely still active today. Additional analysis shows that they are not merely software-controlled bots, and manage their online identities in various complex ways. Finally, we argue that if it is possible to discover these accounts using externally-accessible data, then the platformswith access to a variety of private internal signals-should succeed at similar or better rates.

Characterizing Twitter Users Who Engage in Adversarial Interactions against Political Candidates

“…Note that it is possible that organized information operations took place on Twitter during our data collection period, whether using bots powered by automated algorithms [16,25] or through state-operated campaign [6,34]. We compared our data with a list of state controlled accounts with over 5,000 followers published by Twitter [43] (as the information of this set of accounts was not anonymized): none of them showed up in our dataset.…”

TexTrolls: Identifying Russian Trolls on Twitter from a Textual Perspective

“…After the 2016 US elections, Twitter has detected a suspicious attempt by a large set of accounts to influence the results of the elections. Due to this event, an emerging research works about the Russian troll accounts started to appear [7,32,19,13,2].…”

“…There is a probability that Twitter has missed some IRA accounts that maybe were less active than the others. Based on this hypothesis, the work in [19] built a machine learning model based on profile, language distribution, and stop-words usage features to detect IRA trolls in a newly sampled data from Twitter. Other works tried to model IRA campaign not only by focusing on the trolls accounts, but also by examining who interacted with the trolls by sharing their contents [1].…”

“…A previous work [19] showed that such suspicious accounts are not bots in a strict sense and they argue that they could be considered as "software-assisted human workers". According to [9], the online suspicious accounts can be categorized into 3 main types: Robots, Cyborgs, and Human Spammers.…”

Down the bot hole: actionable insights from a 1-year analysis of bots activity on Twitter

“…Other studies focused on the detection of collective and inauthentic behaviors of malicious accounts to uncover coordinated campaigns [45,38,39] and suspicious content diffusion [31,26,60]. However, social media abuse persists and the online ecosystem still presents a mix of organic and malicious users [34,32], where the former class still demonstrates a moderate capability to identify the latter [56]. This also calls for a clear understanding of users' susceptibility to the content shared by malicious accounts and the interplay with them.…”