Botnet Attack Classification with Deep Learning Models

Abstract

Most devices, ranging from big data servers to the IoT in household device, are accessible on the Internet. Their functioning almost partly or wholly depends on the Internet connectivity. These innumerable devices on the Internet are opportunity for the attackers, to compromise any device and use it for their own benefit. Botnets attack detection and identification is an important security concern. In this paper, we apply various deep learning algorithms to UNBS-NB 15 dataset for the classification of botnet traffic. From our analysis, we understand that deep learning algorithms with their intricate structure are able to perform better than the machine learning algorithms in botnet detection. Also, the hybrid models have an added advantage of canceling out the negatives of the parent algorithms and harnessing only the positives.