Everyone who uses Twitter has been confronted with spam and bots in some form or another, even after the company’s recent pre-IPO cull.
The problem with Twitter’s 140-character post system is that this brevity can make it very difficult for filters to root out bot accounts like email clients root out spam mail. In fact, up until now, Twitter’s policy for removing bot accounts has largely been reliant on reports by real users.
This just wasn’t enough for Imperial College London researchers Gabriela Tavares and Aldo Faisal, who have developed an algorithm which makes it possible to tell with 85% accuracy whether an account is real or a bot. The pair published a study which analyses the timestamps on 165,000 to determine the difference between human activity and spam accounts. If developed further, this algorithm could be applied to become a filter for spam activity on Twitter.
The solution isn’t as simple as what time tweets are posted at, according to researcher Tavares, and it would be incredibly hard for bot accounts to mimic human behavior without having prior knowledge of the algorithm parameters. Aside from the ability to tell regular Twitter users from spammers, another interesting result of the algorithim is the ability to predict the timings of a human’s next tweet with almost pinpoint accuracy.