http://lattes.cnpq.br/6443528495315117; SANTOS, Miqueas Galdino dos.
Resumen:
Advances in microelectronics have allowed the rise of Wireless Sensor Networks (WSNs),
which are increasingly present in our daily lives as a fundamental element of the Internet of
Things paradigm. In this environment, the reliability of the data that transits this network
is a relevant factor that generates investigations and research in the academic environment.
Due to the several limitations existing in the WSNs architecture, sensor failures are common,
generating incongruous and abnormal data. However, abnormalities also reflect changes in
the phenomenon being monitored by the sensors, thus creating problems in defining what
is really happening in a given sensor. Thus, anomalies are indicative that something nonstandard
occurs in the network, and knowing the cause of these abnormalities is essential
for decision-making in the environment. In view of this context, the present work develops
an approach for detecting and categorizing anomalies in wireless sensor networks based on
fuzzy logic, which aims to help determine the existence of events or faulty sensors. Contexts
of different types of data failures were evaluated and what is their relationship with factors
related to the number of failed sensors in a region and packet loss. The results pointed
to the effectiveness in the identification of abnormalities and categorization of anomalies,
with greater effectiveness in the categorization of intermittent failures, in relation to gradual
anomalies and events. It was also found greater effectiveness for environments with fewer
faulty sensors and a moderate relationship was noticed in relation to approach and the loss
of packets in the environment.
