Dinâmica da digitação aplicada à autenticação periódica de usuários em ambientes virtuais de aprendizagem

The authentication of users in Virtual Learning Environments (AVAs) usually requires a password to connect to the environment. Such method is unable to ensure the authenticity of every user activity. In order to mitigate this problem, Cruz et al. proposed an engine to execute periodic and non-intrusive authentication of users in VLE [1]. Machine learning techniques were applied to build recognition models based on the keystroke dynamics of users and it is VLE independent. In order to demonstrate its practical feasibility in a real scenario with a large number of users, this paper conducted an applied case study, using Moodle, to a group of 307 users, producing a total of 4,829 evaluated strings. About 89% of the recognition models achieved at least 80% of accuracy, indicating the effectiveness of the engine at identifying suspicious authorships.