be-nz logo
Story image

New detection tool aims to catch Twitter bots in real time

21 May 2020

Twitter bots are everywhere – sometimes it can be difficult to tell if a tweet is from a genuine person, or some kind of bot. 

That’s why NortonLifeLock Research Group developed a tool that can detect bots in real time to help keen Twitter users filter out the genuine tweets from junky bot tweets.

Bots are becoming notorious for spreading misinformation and many social media platforms like Twitter are trying to stamp out bot accounts. But Twitter recognises that it’s hard to tell just how much misinformation is out there on social media.

NortonLifeLock researchers developed a machine learning model that can detect twitter bots with a ‘high’ degree of accuracy. That model was then developed into the BotSight tool, which is now in beta for web browsers and on iOS.

“To determine whether an account is a bot, we look at over 20 different distinguishing features per case, including the amount of randomness in the Twitter handle, whether the account is verified, the rate at which it is acquiring followers, and the account’s description. We verified our approach by observing BotSight in action,” explain the researchers. 

“Using BotSight’s classifier on what we believe is the largest archive of Twitter’s historical data ever collected outside Twitter (over 4TB), we found many interesting and surprising things. One is that the problem of disinformation is not as small as Twitter’s numbers suggested on first blush, but also nowhere near the more sensational headlines we’ve seen.”

Researchers found that around 5% of tweets come from bots, although that number is dropping.  COVID-19 was an exception, where bots accounted for as much as 18% of all tweets.

“A random sample of the Twitter stream indicates 4-8% bot activity by volume over the same time period. This contrast shows that bots are strategic about their behaviour: favouring current events to maximize their impact.”
 
Researchers say that results differ when accounting for time of day, topic, and language. The entire research process took more than six months so the team could test and improve the machine learning model, as well as the tool’s design.

“It has also enabled us to better understand bots, contextualizing where they are likely to appear and how they act,” researchers state.

The BotSight tool works across Twitter search, trending topics, and users’ home timelines.  

BotSight is still a research prototype beta. Researchers are inviting people to try the tool out and to provide feedback.