Context-Aware Drone Detection

Current commercial and research solutions for drones' detection do not make any assumption on the scenario deployment, as well as the unique mobility pattern associated with the drone's trajectory. Indeed, drones' trajectory is different from the one of people moving at the ground level, being independent of roads layout and obstacles on their path: drones fly directly towards their target, minimizing the travel time and the possibility of being detected. Grounding on this intuition, we propose CADD, a solution enabling drone detection via context-related information. CADD leverages a sensing infrastructure to locate and track all the devices in the area to be protected, and it distinguishes the trajectory of a drone as an anomaly with respect to a ground-truth of allowed trajectoriesDthe ones generated by the devices at the ground level, belonging to vehicles and users within them. We evaluated the performance of CADD over a real dataset of moving vehicles (taxi) in both urban and rural scenarios, resulting in an overall accuracy of 0.91 and 0.84, for the rural and the urban scenario, respectively. The performances of CADD confirm the effectiveness of our solution and show its promising potential for context-related drone detection.