SNEAKY SOFTWARE Fully Automated Propaganda

Millions of so-called bots are busy working to shape public opinion and throw elections. Created at the behest of the US military, they are now easy to deploy - and much harder to fight.
Armies of robot propagandists are becoming ever-harder to distinguish for humans on social media. Photo source: Getty

Konrad Traupe knows something's off when a questionable idea suddenly seems enormously popular. Someone started a petition on the openPetition platform calling for a law prohibiting journalists from using words like "Islamist." Within 10 days, the number of signatures rose from 16,000 to more than 30,000. Mr. Traupe, who works for the platform, suspected robot software of automatically adding thousands of false signatures.

This particular attack came in three waves. First, bots faked signatures on the German-language side of openPetition, then they hit the Turkish-speaking site. Later they tried it on the English version.

Whenever bots attack a petition, Mr. Traupe and his colleagues at openPetition's offices in Alexanderplatz, Berlin have the sisyphean task of combing through the lists for counterfeits. "This is a huge effort for our small team," Mr. Traupe told Handelsblatt. He assumes supporters of Turkish President Recep Tayyip Erdogan are responsible for the meddling bots.

Social bots have the potential to undermine trust in democracy. Bunedstag Office for Technology Assessment

Now, with German national elections on the horizon this fall, experts are worried that extremists, and perhaps foreign governments, are gearing up to manipulate public opinion using bots. The Office for Technology Assessment at the Bundestag, Germany’s parliament, has even warned that "social bots have the potential to undermine trust in democracy."

The closer the election gets, the busier the bots become, writing populist comments on social networks like Twitter and Facebook. The accounts @draco333999 and @saul43, for example, spread identical news of Russian Prime Minister Dmitry Medvedev's criticism of German Chancellor Angela Merkel. According to Tabea Wilke, a digital communications adviser to politicians and founder of the bot information platform, 10 to 25 percent of the related tweets during political TV appearances are from bots.

But how are bots taking over the internet, and can anything be done to stop their invasion?, openPetition's US competitor, has also seen attempts to manipulate petitions in Germany and suspects bots were used. At openPetition, Mr. Traupe and his colleagues are trying to track down the source of the bots by their IP addresses, which can narrow down a user’s physical location. But the false signatures seem to come from addresses scattered across the globe.

That’s because the attackers have passed signals through several different computers around the world. Bot operators can also use services such as VPN and IP obfuscators to disguise their origin. That makes it almost impossible to figure out who is behind the manipulation - and very difficult to defend against it.

According to researchers at the University of Southern California, up to 48 million bots from around the world are present on Twitter.

The public first became aware of the effect of political bots during the Crimean conflict in 2014, when a whole army of bots loyal to Russian President Vladimir Putin busily spread propaganda on social media networks. This model for spreading opinions was developed at Harvard University. Journalist Glenn Greenwald discovered this irony through research into the US National Security Agency.

In a 2008 paper entitled “Conspiracy Theories,” Cass Sunstein, a law professor and adviser to former president Barack Obama, proposed the spread of conspiracy theories through social media networks to weaken foreign governments. He said it was possible to change the minds of a large number of people by introducing an alternative view of things. The US military commissioned California-based HBGary to develop software for this purpose, based on documents published by the hacker group Anonymous in 2011. This software would generate a multitude of Twitter accounts with false identities. And so the robot propagandists were born.

In the meantime, the tools of a select group of hackers have become available to the masses. Any casual programmer can access the building blocks for bots on the web, Tim Berghoff, a security expert at anti-virus software producer G-Data, told Handelsblatt. Defense against this tactic, however, is more difficult.

Calls by German politicians to ban the bots or to punish them as “digital trespassers” are seen as heavy-handed and ineffective. And there are only a handful of people in Germany who truly understand the issue.

One of these experts is Simon Hegelich, professor of politics at the Technical University of Munich. Across from Mr. Hegelich's office, in a stuffy room, is the core of his work: three large Dell computers sit on the floor, with an artfully positioned array of credit card-sized mini computers. These computers search social media networks day and night, looking for discussions on specific topics, such as the influx of refugees. Over the past few months, they have compiled a terabyte of data.

Mr. Hegelich and his team are like criminologists. They look for behavioral patterns that might betray the presence of bots: hyperactive accounts, for example, that send more than 1,000 short messages per day, a pace even Donald Trump couldn't compete with; news stories sent out at a regular pace are also suspicious; or profiles that are just as active on the weekends as during the week.

It's a cat-and-mouse game. These days, programmers give their bots human habits. They have them take a break at night and send messages at irregular intervals. The most advanced use artificial intelligence and can respond to tweets as if they were a person.

"We won't be able to prevent the manipulation, but we can throw light on it," said Mr. Hegelich. He will soon publish the results of his analysis and hopes fewer people will allow themselves to be fooled. Real power over the spread of these stories, however, is in the hands of the operators of platforms like Facebook and Twitter.

These companies have more money, more staff and more data to hunt down the bots. But these same bots support their business model. Higher user numbers mean companies can charge more for advertising. Facebook, at least, does now block the bots it identifies. Twitter, however, isn’t doing as well as a business, and every user counts.

According to researchers at the University of Southern California, up to 48 million bots from around the world are present on Twitter. Early Twitter investor Chris Sacca recently said the ongoing bot issue was “embarrassing” for the company. For their creators, though, they are undeniably effective. These days, bots are involved in every major debate in Germany, Mr. Hegelich warns. And they look set to operate largely unchallenged throughout the German elections.


This article originally appeared in German weekly business magazine WirtschaftsWoche. To contact the author: [email protected]