A Hybrid Approach for Anomaly Detection in the Internet of Things

Internet of Things is the next generation of internet that physical things is going to interacts with together without human interventions. Its presence in different domains and environments, has improved the quality of human lives. This emerging technology due to the limited sensor's resources and environmental influences is prone to the existence of abnormal data; in addition, due to the distributed characteristic and its heterogeneous elements, applying complex anomaly detection techniques is very difficult. In this paper, a hybrid approach based on K-Means clustering and Sequential Minimal Optimization (SMO) classification has been presented for anomaly detection in the IoT which despite of very low complexity has a high accuracy in detecting anomalies. This method has tested on the data collected from sensors in the Intel Berkley research lab, which is one of the available free dataset in the domain of IoT. The results shows that the proposed technique could achieve an accuracy of 100%, positive detection rate of 100% and reduce false positive rate to 0%.