Low-rate distributed denial of service attacks detection in software defined network-enabled internet of things using machine learning combined with feature importance

Abstract

One of the main challenges in developing the internet of things (IoT) is the existence of availability problems originated from the low-rate distributed denial of service attacks (LRDDoS). The complexity of IoT makes the LRDDoS hard to detect because the attack flow is performed similarly to the regular traffic. Integration of software defined IoT (SDN-Enabled IoT) is considered an alternative solution for overcoming the specified problem through a single detection point using machine learning approaches. The controller has a resource limitation for implementing the classification process. Therefore, this paper extends the usage of Feature Importance to reduce the data complexity during the model generation process and choose an appropriate feature for generating an efficient classification model. The research results show that the Gaussian Naïve Bayes (GNB) produced the most effective outcome. GNB performed better than the other algorithms because the feature reduction only selected the independent feature, which had no relation to the other features.