Abstract
Nowadays, online social media became an essential communication platform that affects all our lives. However, it goes unnoticed that there are entities found in social media that affect and manipulate society and pose as a human to do so; these entities are called bots. Over the years, researchers have conducted tremendous work in developing bot detection applications, but unfortunately, most of them focus on analyzing and detecting English bots, whereas there is a noticeable lack in addressing Arabic bots. Therefore, in this paper, we have studied and analyzed Arabic Twitter bots' behavior, and highlighted the most significant features of these bots. In addition, we addressed the language and cultural effect on these features and whether the bots' behavior changes based on language and culture. The study started by collecting datasets using crowdturfing campaign and manual labeling. Then we extracted the features that were identified and proven to be useful by other researchers. After that, we built a classifier that tested the probability of an account being human or bot. Finally, the results were analyzed and discussed.