Social networks can quickly propagate information to large audiences and can be used to spread fake news or to provide false figures of popularity. Social bots, i.e., software robots that automatically interact with human users and produce content under a fictive identity, are used for such harmful activities. In this paper, we study the relationship between bots and genuine human users with the aim of identifying those “credulous” users who are particularly exposed, and unintentionally contribute, to the activities planned by a network of bots. Spotting credulous users is useful to service providers to display warnings on their dashboards, scan their activities for early signs of attacks, or take more active measures to prevent or limit the negative effects of their activities. Here we aim at identifying credulous users on Twitter starting from those involved in any social relation with a bot. To that end, we rely on an existing bot detector along with its dataset of genuine users and bots that we extend with additional information about the friends of each genuine user. To single out credulous users out of genuine ones, we study the effectiveness of different metrics or combinations thereof. We see this as a first step towards singling out features that can be used to detect credulous users without resorting to the expensive analysis of the nature of their friends.
Identification of credulous users on twitter
Balestrucci A.;De Nicola R.;
2019-01-01
Abstract
Social networks can quickly propagate information to large audiences and can be used to spread fake news or to provide false figures of popularity. Social bots, i.e., software robots that automatically interact with human users and produce content under a fictive identity, are used for such harmful activities. In this paper, we study the relationship between bots and genuine human users with the aim of identifying those “credulous” users who are particularly exposed, and unintentionally contribute, to the activities planned by a network of bots. Spotting credulous users is useful to service providers to display warnings on their dashboards, scan their activities for early signs of attacks, or take more active measures to prevent or limit the negative effects of their activities. Here we aim at identifying credulous users on Twitter starting from those involved in any social relation with a bot. To that end, we rely on an existing bot detector along with its dataset of genuine users and bots that we extend with additional information about the friends of each genuine user. To single out credulous users out of genuine ones, we study the effectiveness of different metrics or combinations thereof. We see this as a first step towards singling out features that can be used to detect credulous users without resorting to the expensive analysis of the nature of their friends.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.