This just in: Hillary Clinton leads the Twitter zombie vote! She is closely followed by Jeb Bush and Donald Trump. With more primaries on the horizon, the zombie swing vote could be the edge Hillary needs to win the nomination!
So What Exactly is a Twitter Zombie?
It’s no secret that Twitter is littered with automated accounts, also known as bots. According to their own SEC filing, as many as 8.5% of monthly active users operate “without user-initiated action.” In total, there may be as many as 42.5 million automated accounts.
Unfortunately, the majority of bots aren’t so benign. Some distribute spam and malicious links. Some promote pirated material and counterfeit goods. Some even act as command and controls for malware, DDoS and other cyber attacks launched or coordinated via Twitter.
But a Twitter zombie account is something different — these accounts are built in bulk with the purpose of inflating follower counts. These are the giant roving social botnets that can be purchased by shady marketers or others attempting to legitimize their accounts for potentially nefarious purposes.
Because these accounts are built in bulk they are generally sparse, lifeless and quite zombie-like. This means they typically have only a handful of auto-generated tweets and fellow zombie followers. As far as fraudulent accounts go, they are optimized for quantity, not quality, making them resemble a pack from The Walking Dead.
So Who Wins the Zombie Demographic?
Naturally, being the data science nerds we are, ZeroFox recently decided to find out which presidential candidate had the largest Twitter following of zombie accounts. To do this, we built a machine learning algorithm which classifies Twitter accounts as zombies or not zombies. This involves separating zombies not only from real, human-run accounts, but from other non-zombie bots as well. In order to do this, we set up a shell Twitter account and purchased fake followers, which came at the bargain price of $30 for 10,000 followers.
Machine learning is a method of classifying things based on a set of features (if I know that cubes have distinct features such as hard edges, corners, and 6 sides, I can classify them as distinct from spheres, which are rounded, have no edges and no corners). Machine learning algorithms can ingest lots and lots of features, and as it is “trained,” it gets better and better at identifying and weighing the features. Eventually, a good machine learning model can classify objects with near perfect accuracy.
For zombies, a few of the most informative features include the number of posts, ratio of followers to friends, and whether or not they use default profile settings. Our final classifier identifies undead Twitter accounts with 99% accuracy.
With classifier in hand, we ran it against a sample of 50,000 of the presidential hopefuls’ followers. The polls are in, and Hillary Clinton takes the cake: our model reveals that a whopping 20.6% of her followers are zombies. Coming in second is Jeb Bush with 12.3% and Donald Trump with 12.1%. Bringing up the rear are Bernie Sanders with 6.9% and Chris Christie with 5.7%. See the full breakdown below.
It’s clear the undead voting block is divided, lumbering equally to each side of the aisle. There are zombie Democrats and zombie Republicans. There are zombie Socialists, Tea Partiers and Libertarians. There are potentially even some zombie Green Party supporters out there. Dawn of the Dead is November 8th, 2016 at which point we’ll see if these Twitter zombies can stumble out to the polls without getting distracted by all the juicy braaaaains.
How did they end up following the candidates? It’s possible that the zombies were purchased by the social media teams of the campaigns, but more likely than not, as anyone can purchase followers, they were bought by a candidate’s supporters. Sometimes, these roving hordes simple latch on to popular accounts with no rhyme or reason.
So Why Do Zombies Matter?
The potential implications of Twitter zombie followers are numerous. The main objective? Deception. An inflated follower count makes the candidate look more popular than they actually are, causing the everyday user to misinterpret the neck-and-neck jockeying for supporters. Certainly, a horde of brain-dead followers won’t translate to a guaranteed win at the polls.
In addition, an impostor account which is perceived to be legitimate due to its high follower count, can infuse counter productive or negative sentiment into a hotly debated topic.
Perhaps most worrisome, especially if the campaign social media managers don’t know where the zombies originated (viral epidemic? voodoo? witchcraft?), they can become malicious without a moment’s notice. A Twitter zombie is defined as a husk of an account — very few posts, followers or account details — but can easily begin posting scams or phishing links if whoever controls the bots, chooses to do so. We see zombie accounts turn particularly vicious all the time (imagine a Night of the Living Dead zombie turning into a 28 Days Later zombie). At the click of a button, these groaning, shuffling brain-eating automatons could take off at full sprint, posting malware and target legitimate followers.
Stay safe while surfing Twitter, and be aware of both the living and the dead.