Evaluation of Network Intrusion Detection Systems for RPL Based 6LoWPAN Networks in IoT

Over the past few years, Internet of Things security has attracted the attention of many researchers due to its challenging and constrained nature. Particularly in the development of Network Intrusion Detection Systems which act as first line of defence for the networks. Due to the lack of reliable Internet of Things based datasets, intrusion detection approaches are suffering from uniform and accurate performance advancements. Existing benchmark datasets like KDD99, NSL-KDD cup 99 are obsolete and unfit for the evaluation of Network Intrusion Detection Systems developed for RPL based 6LoWPAN networks. To address this issue, the RPL-NIDDS17 dataset has recently been generated. This dataset consists seven types of modern routing attack patterns along with normal traffic patterns. In the proposed dataset we consider twenty two attributes that comprise of flow, basic, time type of features and two additional labelling attributes. In this study, we have shown the effectiveness of RPL-NIDDS17 by statistically analysing the probability distribution of features, correlation between features. Complexity analysis of the developed dataset is done by evaluating five machine learning techniques on the dataset. Evaluation results are shown in terms of two prominent metrics accuracy and false alarm rate, and compared with the results of KDD99, UNSW-NB15, WSN-DS datasets. The experimental results are presented to show the suitability of our proposed RPL-NIDDS17 dataset for the evaluation of Network Intrusion Detection Systems in Internet of Things.