03 A Multi-agent Approach for Online Twitter Bot Detection

Online social networks are tools that allow interaction between human beings with a large number of users.Platforms like Twitter present the problem of social bots which are controlled by automated agents potentially used for malicious activities. Thus, social bot detection is important to keep peop...

Descripción completa

Detalles Bibliográficos
Autores: Fonseca Abreu, Jefferson Viana, Ghedini Ralha, Célia, Costa Gondim, Joäo José
Tipo de recurso: capítulo de libro
Fecha de publicación:2021
País:España
Institución:Universidad de Castilla-La Mancha
Repositorio:RUIdeRA. Repositorio Institucional de la UCLM
OAI Identifier:oai:ruidera.uclm.es:10578/28601
Acceso en línea:http://doi.org/10.18239/jornadas_2021.34.03
http://hdl.handle.net/10578/28601
Access Level:acceso abierto
Palabra clave:Bot detection
Social bots
Twitter
Multi-Agent System
MAS
Descripción
Sumario:Online social networks are tools that allow interaction between human beings with a large number of users.Platforms like Twitter present the problem of social bots which are controlled by automated agents potentially used for malicious activities. Thus, social bot detection is important to keep people safe from harmful effects. In this work, we approach the Twitter bot online detection problem with a multi-agent system (MAS). It is based on supervised classification with three machine learning algorithms and a reduced set of features. The MAS performance compared to the three algorithms applied separately - Random Forest, Support Vector Machine, and Na¨ıve Bayes - presented similar results. Besides, interesting results for online bot detection with the MAS prototype suggested that 88.19% of bots detected were correctly labeled. The results indicate that the approach used is feasible and promising for the real-time bot detection problem.