Drones' Face off: Authentication by Machine Learning in Autonomous IoT Systems

Autonomous Internet-of-Things (IoT) are comprised of moving objects such as drones and rovers that use self-control techniques to accomplish a mission while following a path. However, losing control in such systems usually by spoofing their sensors or hijacking with misleading commands can lead to catastrophic safety consequences. In this paper, we close the gap by authenticating the behavior of autonomous IoT systems during operation. In particular, we check the behavior of a moving IoT object, e.g., a drone, by evaluating its time-series telemetry traces during the flight. We examine three different machine-learning algorithms for this purpose, namely, K-Nearest Neighbour (KNN), Support Vector Machine (SVM), and Logistic Regression (LR). Our results show that KNN is the best method of the three selected techniques for authentication in dynamic IoT systems, e.g., drones. We achieved 93.4% in precision rate and 100% recall rate with KNN.