LAUSR

IoT networks mostly rely on wireless mediums for communication, and due to that, they are very susceptible to intrusions. And due to the tiny nature, processing complexity, and limited storage capacities, IoT networks require very reliable intrusion detection systems (IDS). Although there are many IDS types of research available in the literature, most of these systems are suitable for wired network environments, and the benchmark datasets used for these research works are mostly relying on wired datasets such as KDD Cup’99 and NSL-KDD. IoT and wireless networks are distinct in nature as wireless networks give more emphasis on the data link layer and physical layer. These concerns are not given much attention in traditional wired datasets in the body of knowledge. Therefore, in this research, an IDS system is developed using a newly available IoT wireless dataset (NaBIoT) in the literature with the datasets focusing much on the common IoT related attacks, and related layers are taken into consideration. The IDS system developed is evaluated by comparing with various machine learning algorithms in terms of evaluation metrics such as accuracy, F1 score, false positive, and false negative. Moreover, the IoT wireless dataset is compared against the traditional NSL-KDD datasets to evaluate the need for IoT wireless datasets. The NaBIoT datasets show its effectiveness in detecting wireless intrusions. Besides that, the simulation is performed with different combinations of features to conclude that certain features are primary in detecting attacks, and IDS does not require all the features to perform detection. This can reduce the detection time mainly for machine learning and creating the models. This research results have proposed some of the critically important features to be used and eliminating not such important features.