Bots sociais: implicações na segurança e na credibilidade de serviços baseados no Twitter

Abstract

More and more, data extracted from social networks is used to build new applications and services, such as traffic monitoring platforms, identification of epidemic outbreaks, as well as several other applications related to the creation of smart cities, for example. However, such services are vulnerable to attacks from bots - automatized accounts - seeking to tamper statistics of public perception posting an excessive number of messages generated automatically. Bots can invalidate many existing services, which makes it crucial to understand the main forms attacks and to seek defense mechanisms. This work presents a wide characterization of the behavior of bots on Twitter. From a real data set containing 19,115 bots, several characteristics of bots were identified, extracted from behavior and writing patterns, that have discriminative power. From these features, we present an automatic detection method capable to detect 92% of the bots while only less than 1% of real users are misclassified. In addition, we conducted a study on which characteristics makes a bot most successful in infiltration tasks. For this study we created 120 socialbots on Twitter. During 30 days we monitored their behavior and interactions with all network users, as well as 600 target users. During this period our bots had 5,966 interactions with 2,637 Twitter users.