Communications of the ACM

how to discriminate between bots and humans. User meta-data is considered among the most predictive feature and the most interpretable ones. 22, 38 We can suggest a few rules of thumb to infer whether an account is likely a bot, by comparing its meta-data with that of legitimate users (see Figure 2). Further work, however, will be needed to detect sophisticated strategies exhibiting a mixture of humans and social bots features ( sometimes referred to as cyborgs). Detecting these bots, or hacked accounts, 43 is currently impossible for feature-based systems.

Combining Multiple Approaches

Alvisi et al. 3 recognized first the need
of adopting complementary detec-
tion techniques to effectively deal
with sybil attacks in social networks.
The Renren Sybil detector37, 42 is an
example of system that explores mul-
tiple dimensions of users’ behaviors
like activity and timing information.
Examination of ground-truth click-
stream data shows that real users
spend comparatively more time mes-
saging and looking at other users’
contents (such as photos and videos),
ROC via cross validation. In addition to
the classification results, Bot or Not?
features a variety of interactive visual-
izations that provide insights on the
features exploited by the system (see
Figure 1 for examples).

Bots are continuously changing and evolving: the analysis of the highly predictive behaviors that feature-based systems can detect may reveal interesting patterns and provide unique opportunities to understand

Figure 2. User behaviors that best discriminate social bots from humans.

Social bots retweet more than humans and have longer user names, while they produce fewer tweets, replies and mentions, and they are retweeted less than humans. Bot accounts also tend to be more recent.

No. retweets No. tweets No. replies No. mentions No. times retweeted Username length

Z-score

– 3 – 2 – 1 0

Human

1234

Account age
Social bot

Figure 1. Common features used for social bot detection. (a) The network of hashtags co-occurring in the tweets of a given user. (b) Various sentiment signals including emoticon, happiness and arousal-dominance-valence scores. (c) The volume of content produced and consumed (tweeting and retweeting) over time.