On 17 April 2018, Starbucks announced it would close 8,000 US stores for an afternoon, to allow its 175,000 employees to undergo racial-bias training. The training was an attempt to allay the public outrage over two black men being arrested for “trespassing” while they were waiting for a friend inside a Philadelphia store the week before. The day after the announcement, counterfeit coupons promising free coffee as an apology started making the rounds on social media. The coupons, which were available exclusively to people of colour, could be traced back to internet trolls from the anonymous message board 4chan.
This was not the first time Starbucks was the target of a coordinated online campaign and suffered reputational damage, but it was a situation that could have been handled better. “The relatively minor action taken suggests they didn't understand the true scale of the fake campaign coordinated against them,” says Ali Tehrani, the cofounder of London-based startup Astroscreen. Tehrani’s company uses machine learning and human analysts to scour Twitter and Facebook and spot fake accounts and disinformation campaigns against governments and brands. It then alerts its customers when they are under attack, allowing them to devise a strategy to defend their reputation online.
In the case of Starbucks, the company only set the record straight in response to media queries and did not release a statement on their website or social media accounts. The lack of communication backfired and gave trolls an opportunity to put out a false company statement, fuelling the hoax cycle and confusing customers even more.
Social networks have changed the way we consume and share the news, but are also easy to manipulate for commercial and political gain. Astroturfing – an online manipulation campaign that is not automated but uses fake accounts and groups to create the impression of a widespread grassroots movement in support of or in opposition to a brand or political message – is hard to tackle. “The problem is that you can't just build one magic algorithm to ban all fake accounts. A lot of these accounts are hard to distinguish from real users,” says Tehrani.
The company has developed techniques to detect social media manipulation based on a topic of interest – during an election campaign, for instance, this could be the names of politicians, policy issues or specific hashtags and events.
Malevolent actors might create thousands of Twitter accounts to interfere with a political topic but, because they aren’t able to manually come up with that many unique usernames, they will likely use words and numbers that follow a certain pattern Astroscreen can recognise. The timing and frequency of the posts will also indicate whether the accounts are authentic or not. When Astroscreen’s automated system picks up suspicious online activity, the company’s disinformation experts step in and use linguistic analysis to discern if a single person is behind multiple accounts based on the way they write.
Tehrani previously founded a news analytics company, which he sold in 2016, shortly before Donald Trump’s election put the problem of fake news under the spotlight. “I had this really strong feeling that the problem with disinformation isn’t fake news. In fact, it's just the manipulation of social networks to amplify one message over another,” says Tehrani.
His team takes on a similar role as Twitter and Facebook’s own content moderators, which sift through posts that have been flagged by users or the social network’s AI tools. “Right now, Twitter and Facebook decide what topics to monitor with their analysts and they don’t have many of them,” says Tehrani.
In a way, Astroscreen does for social media what cybersecurity software did for Microsoft when it launched more than 40 years ago. “In the early days, people assumed Microsoft should build secure operating systems, but very quickly they realised that they need an additional layer of security,” says Tehrani. That’s when the cyber security industry was born. Tehrani says that his company is working with businesses and governments to monitor coordinated online campaigns across all major social networks and discussion websites such as Reddit.
Spotting targeted attacks as early as possible gives organisations an opportunity to react and engage their lawyers and communications professionals, and if needed, report the activity to the relevant social networks. In some cases, the organisation under attack might decide to do nothing as any proactive outreach might give harmful content more attention that it needs. “We just want to give them guidance and a report in a timely manner with enough context to help them make the best judgement call,” says Tehrani.
This article was originally published by WIRED UK
