ANKARA
At least 47% of the Twitter trends in Turkey are fake, and 20% of those trends make up worldwide trends, according to a study by researchers in Switzerland.
Since 2015, hit-and-run astroturfing attacks – using automated accounts (bots) to artificially propel a chosen keyword to the top of Twitter trending topics – have hit Turkey's 11.8 million active users and global trends.
When determining trends, Twitter does not consider whether a tweet has been deleted, making it more vulnerable to such attacks.
Tugrulcan Elmas, co-author of the 2019 study Ephemeral Astroturfing Attacks: The Case of Fake Twitter Trends, told Anadolu Agency those attacks "undermine the integrity of the social media platform."
"If trending topics fail to show what is actually trending, the users will lose trust in the mechanism and maybe the platform as a whole," said Elmas, a doctoral student at the Swiss Federal Institute of Technology in Lausanne.
"The high volume of inorganic activity and the company’s lack of measures to prevent it will also affect its reputation," he warned.
Users are not only exposed to illicit ads, hate speech, phishing apps and topics that are irrelevant to what is trending, he said, but also see instances of the media and politicians addressing topics because they are trending on Twitter.
Some academic studies even cite trending topics as a proxy for public opinion, apparently unaware that such trends can be manipulated rather than organic.
"One example is #SuriyelilerDefolsun, as we found it was cited in at least five academic studies,” said Elmas, referring to a hashtag supposedly showing support in the Turkish public for Syrians who fled the civil war to leave Turkey.
“But all [of them] attributed the source of the campaign to social media users, missing the bot activity."
The Turkish government and society have been praised internationally for welcoming Syrian refugees, some 4 million of whom live in Turkey, more than any other country.
Bot accounts can create topics from scratch
Elmas confirmed that bot automated social media accounts even have the ability to create topics from scratch.
"Trending topics are the topics that are popular at a given time and are updated every five minutes," he explained. "Coordinated and bursty bot behavior that gives the algorithm the impression of organic activity can create topics from scratch."
In 2019, the study found, there were over 6,500 fake trends in Turkey, with the largest group (32%) being illicit advertisements, followed by people seeking policy changes (18%), politics (12%), and "cult slogans" (9%).
Twitter knows about attacks
Even though Twitter may see fighting these attacks as difficult, said Elmas, "It doesn’t matter how hard it is to fix; the fact is that a huge proportion of trends are fake, and that should be handled."
He said they notified Twitter in July 2019 after encountering the flaw in the algorithm, and the company acknowledged that the attacks do exist, saying: "We think it is a valid security issue and we will look into that."
"After a month or so, they replied again, thanking us, stating that it’s about their spam filters and they are currently satisfied with their spam filters and are not going to introduce any changes," he added.
He noted, however, that a year later Twitter suspended 7,340 so-called troll accounts involved in such attacks.
"However, the attack was, and is, still ongoing, since banning some of the accounts involved does not fix the vulnerability itself," Elmas explained.
"The data Twitter published also included some of the fake trends that we uncovered in our work and bot tweets promoting these fake trends."
When the researchers contacted Twitter for a second time in June 2020, the social media platform said they would forward their study to the relevant team to address, according to Elmas.
"We have followed up since, but have not received any indication that they are progressing," he said.
Hitting back
The study suggested there are two options to prevent hit-and-run astroturfing attacks that manipulate trends.
A detection mechanism could be implemented to prevent malicious tweets from being considered for trends or even made visible at all, and once a trend is found to be manipulated, it could be removed from trends or even prevented from ever reaching them, it said.
Or Twitter could render the attack useless by allowing the algorithm to account for deleted tweets when computing trending topics, according to the study.
