Internet of Things, the connection of objects such as computing machines, embedded devices, equipment, appliances, and sensors to the Internet, are generating a huge quantity of data in real time. Many of these devices connected to the internet can be used for the generation of spam emails, which can affect both companies and individual users. Because of that, e-mail servers need to be updated to detect in an efficient way these messages. To analyse these emails, there have been proposed various approaches using traditional data mining learning algorithms. But these algorithms require having the data previously stored in order to classify them. Due to the temporal dimension of the data and the dynamism of these spam emails, the target function to be learned can change over time, a problem commonly known as concept drift. In this paper we propose an approach to filter spam mails based on online ensemble classifiers. The predictive performance of these algorithms is evaluated on the benchmark corpus constructed in this work. The experimental results show that that online ensemble algorithms can be an efficient alternative for the e-mail spam detection.