In a years-long awaited move, Twitter announced that it will start fighting off the vast armies of trolls poisoning the beloved social network.
This is badly needed, as I saw last summer, when I was following the #MH17 hashtag where in a matter of hours armies of Moscow-paid trolls started poisoning the hashtag feed with the most toxic, most poisonous, most fake allegations.
Other than reporting accounts as spam, I, as a regular Twitter user had little leverage into voicing down the obviously fake accounts and into helping the world by distributing verifiable, verified, quality information.
At some point, it has gotten so bad that trolls were saying that the missing MH370 flight (en route to Beijing) was kidnapped by USA and/ or NATO and destroyed, months later, above Eastern Ukraine and that MH17 (downed by Russian military in Ukraine) was just some elaborate hoax meant to blame it on the Russians.
This made me sick to my stomach, with so dirty allegations while so many people died and with so many grieving people left behind, so many untold goodbyes and, yet, an army of brainless, spineless, for-profit online thugs were trying really hard to hide the truth, or the graphics of their actions of looting and so much more.
So, dear Twitter, your phone validation of Tor-accessed accounts is a good thing. But it does not suffice so here I come to your helping, expressing some ideas that might help you identify trolls.
1. Use text patterns
Let’s take, for instance, the #Ukraine hashtag. Most trolls use words such as “junta” to describe the democratically-elected government of Ukraine. As a matter of fact, these trolls received a series of memes to be spread about Kiev and were trolling using lists of tweets that were already provided to them, with little or no modifications whatsoever.
2. Use behavioral patterns
As with #MH17, the feed was intoxicated by new accounts, or by accounts that were left inactive for months at a time, none of them verified.
Also, they were arguing, mostly, with unverified accounts, and were attacking victims in hounds of fake/ new/ unverified accounts. This led to their victim getting overwhelmed with answering the bullsh*t allegations, rendering it unable to disseminate any more quality information.
They were effective.
3. Use information sources patterns
More often than not, in regards to #MH17, these trolls were spreading fake information by retweeting questionably-valid information, like that produced by RT. All, all trolls I ever encountered on #MH17 and #Ukraine are actively spreading RT and TASS propaganda.
So, it some new/ inactive/ unverified users are suspiciously-eager to distribute information coming from questionable sources, maybe their intents are questionable.
4. Use hashtag-specific patterns
When something important happens, a hashtag is generated by the community and trolls flood the discussion. It is the discussion, however, that’s betraying their actions: trolls that never spoke of some subject, not even remotely, were becoming very active and very proficient on the topic at hand.
5. Use language-related patterns
What are the odds some Tor-connected fellow in Canada only tweets in Russian? Or Thai? Or only on a topic and only in a certain language? Nill.
I believe these five methods could generate weights. Meaning, some tweet with a weight of 8 is very, very likely to be just trolling and we don’t want that, do we?