Recent years have seen considerable growth for consumer unmanned aerial vehicles (UAVs). Consumer UAVs, while their enormous economic development potential, offer significant security and privacy concerns due to the fact that they enable a wide range of applications. Both for invasion detection and forensics, it is essential that intruding UAVs be quickly detected and identified in order to reduce the dangers. We propose a machine learning-based framework for rapid UAV identification via encrypted Wi-Fi data to supplement the current physical detection methods. Because many consumer UAVs utilise Wi-Fi connections for video streaming and control, the project was inspired by this. Using just the packet size and inter-arrival time of encrypted Wi-Fi data, the proposed system can detect and identify UAVs and their operating modes with ease. Our approach uses a re-weighted '1- norm regularisation to speed up online identification by taking into account the amount of samples and computation costs associated with various characteristics. As a result of this approach, feature selection and prediction performance are simultaneously optimised for a single goal. We use the maximum likelihood estimation (MLE) technique to estimate the packet inter-arrival time in order to deal with packet inter-arrival time uncertainty while optimising the trade-off between detection accuracy and latency. Using real-world Wi-Fi data traffic from eight different kinds of consumer UAVs, we gather a significant amount of data for thorough testing of our new approach.

Unmanned aerial vehicle (UAV) detection, machine learning, encrypted Wi-Fi traffic classification.