Cyber-security firm F-Secure, which makes anti-virus and cloud-protection software, has said that Twitter accounts based outside of the UK have been boosting pro-Leave Brexit content and exhibiting much more "inorganic" behaviour than pro-Remain accounts.

Researchers gathered data from approximately 24 million tweets from 1.65 million users, all of which matched the word “brexit”. Once this foundation was established, F-Secure recorded how many times each user tweeted, retweeted, and how they shared content - through hashtags, URLs, and amplifying other users.

After identifying accounts which both supported the UK remaining in the European Union and those advocating leaving the political bloc, it was revealed that accounts who supported the latter were more likely to:

• Publish hundreds of tweets at certain times during the day, if not constantly.
• Have thousands of followers despite being only a few days or weeks old.
• Retweet more than 99% of the time (behaviour that's indicative of an automated account).
• Have a similar number of followers to other, similar accounts, which could imply that they are controlled by a 'bot herder' and programmed to follow other managed accounts.

F-Secure also pointed out "exceptional behaviours" exhibited by the pro-leave community, which included relying on support from non-authoritative news sources, receiving a disproportionate amount of engagement, and sharing a mixture of Brexit and non-Brexit issues (specifically #giletsjaunes, and #MAGA).

Although this does not necessarily mean that there is a coordinated campaign to manipulate public opinion, F-Secure says it warrants further investigation. The firm admits itself that suspicious activity is difficult to separate from real users, due to the fact that virality can be purchased through online vendors and the only way of tracking that is only possible if those retweets are captured as they are published.

Moreover bot automation is becoming more sophisticated, which means that the usual indications of "artificial amplification" can be hidden. This includes performing actions at random times, and mixing in tweets from Twitter's main website with "share" buttons on news sites to improve realism.