Internet of things (IoT) and Machine to Machine (M2M) technology has become indispensable in everyday life, finding real-time applications in smart tracking/monitoring services, wearable technologies, telemetry, autonomous vehicles, agriculture, healthcare, transportation, etc. It employs millions of sensors and devices connected to automate the tasks to be accomplished by collecting data, activating sensors and communicating between the devices either through the internet or point-to-point connections. Consequently, the smart device usage and the connected devices in the IoT network is increasing exponentially with limitless potential. However, with the increasing connectivity comes the increasing attack surface with attack entry points from different sources in the network. Hence security related risk and network anomalies need to be detected earlier in-order to prevent the devices in the IoT network from bringing down the whole network. In our proposed system, two-layers IoT Stacked Ensemble Learning Network (IoT_SLINK) is designed for attack detection in the IoT network at the source with feature selection using machine learning algorithms. This work focuses on detecting attacks originating from OSI layer 1 to layer 6based on the Random Forest (RF), Decision Tree (DT), Navie Bayes (NB) and K-Nearest Neighbour (KNN)machine learning algorithms. The feature selection algorithms are used to improve the model’s performance in terms of accuracy, reduce overfitting and reduce the training time. The prominent features from the datasets are selected by using Particle Swarm Optimization (PSO) algorithm and Improved Salp Swarm Optimization Algorithm (ISSA) feature selection technique. The performance of the proposed system is evaluated with three IoT benchmark datasets IoT23, BoT-IoT and Distributed Smart Space Orientation System (DS2OS).The proposed system outperforms state-of-art techniques with more than 99% of accuracy with very low false positive rate.

Internet of Things (IoT), Machine to Machine (M2M) technology, attack detection, Machine Learning (ML), Ensemble Learning.