A Hybrid Learning Approach for Automatic Data Labelling and Anomaly Detection in IoT Networks

Internet of Things (IoT) is an environment in which digital equipment is augmented with sensors to share and receive data through network. When devices share data it can be effected by anomalies or any attack due to corrupted data or by any other uncertainty and ambiguity in data. The data can also be corrupted through a damage in device. These attacks or anomalies damage the working of the IoT networks. The anomalous data can be detected through detection techniques however most anomaly detection techniques depend upon labelled data but for IoT datasets, usually class labels are not available. Labeling is performed by a manual process which is time consuming and also costly. As data in IoT increases day by day so there is a need to label and classify data for future unseen data. In this paper a hybrid algorithm is proposed in which both clustering and classification techniques are applied for automatic labeling and classifying on IoT dataset. The model contains two function. In the first phase k-means clustering is employed for labelling dataset instances as normal or anomalous. In the second phase labelled dataset is used to train Random Forest model to detect anomalies in IoT networks. The results show that the proposed model is detecting anomalies in IoT networks with an accuracy 98%, precision 98 %, recall 98%, and F-meausre 0.98%.