Drone accidents have increased Drone accidents have increased Drone - PowerPoint PPT Presentation

Matthan : Drone Presence Detection by y Id Identifying Physical Sig ignatures in in the Drone’s RF F Communicati tion Phuc Nguyen , Hoang Truong, Mahesh Ravindranathan, Anh Nguyen, Richard Han, Tam Vu

Drone accidents have increased

Drone accidents have increased Drone Crashes Through Window Drone Hits an Airplane Drone Hits Man's Head

Ways to take down illegal drones Skywall 100 Drone Defense System Battelle Drone Defender Assume drone presence is known a priori Excipio Anti Drone System Drone Catching Eagles

Existing acoustic-based detection Tien Pham et. al., U.S. Army Research Laboratory

Existing video-based detection Disadvantages: 1. Short range (max. 50m) 2. Require Line of Sight Artem Rozantsev et. al., 2015 3. Light Condition Dependent 4. Hard to differentiate between Drone and Birds Tamas Zsedrovits et. al., 2011, 2012

Existing active radar-based detection This technique creates much interference to the environment and expensive

Can we detect the drone using a Wi-Fi access point? Cost-effective Ubiquitous Internet connected

Matthan Drone Detection System Explore physical signatures in the received RF signal: • Body Shifting (caused by Control Loop Mechanism) • Body Vibration (caused by Propellers Motion) Rx Matthan

Challenges Rx Rx Rx Drone movements can Wi-Fi embedded Wi-Fi hotspot inside happen with Mobile devices moving vehicles unpredicted patterns

Body Shifting Observ rvation ….. Corresponding movement waveform: Idea: The body movement of the drone can be detected by a wavelet transform analysis An example wavelet

Body Shifting Validation • Drone is attached TX antenna and IMU • Observe the signal from IMU and RX when the drone is flying (indoor) The drone movements modulate the wireless signal that sent from the transmitter attached to it

Validating the use of Wavelet Transform to detect body shifting ….. EMF Body shifting signature Take off Noise Landing

Body Vibration Observ rvation Body vibration signature Idea: The body vibration of the drone can be detected by a Fast Fourier Transform analysis

Body Vibration Validation IMU Microcontroller Bluetooth Module IMU data Wireless data

Matthan’s Overview 1 4 2 5 3 6

Evaluation Antenna • Hardware • SDR USRP B200 • 2.4GHz directional antenna • Carrier Sensing: Wi-Fi Analyzer app on Android USRP • Environment setup Laptop • Drones used: Parrot Bebop, Protocol Dronium One Special Setup Edition, Sky Viper, Swift Stream, Parrot AR Drone, Protocol Galileo Stealth, and DJI Phantom • Environments: Urban, Campus, Sub-urban • Distance: 10m  600m 7 types of used Drones

Detecting Different Drones Distance = 50m

Detect Drones at Different Distances 100 96.4 96 97 93 95 92.6 93.9 91.6 95.9 91.3 91 90.4 90.3 89.2 90 92.2 87.5 Percentage 86.4 85.2 84.9 85 87.2 86.6 84.8 83.2 80 81.7 81.5 75 70 10 50 100 200 300 400 500 600 Distance (m) Accuracy Precision Recall Distance from 10m to 600m

Detecting Drones at Different Environments

Drone Differentiation Drone Bebop DJI Galieo Dronium Sky Swift AR Viper Stream Drone Vibration Freq. 60 Hz 100 Hz 140Hz 35 Hz 50 Hz 20 Hz 70 Hz Max Min

Future Works • Develop an automated channel sensing (similar to cognitive radio spectrum sensing) • Integrate automated steering/ beamforming antenna • Localize the position of the drone • Detect multiple drones at the same time

Conclusions • We introduce a system to detect the presence of the drones by identifying unique signatures: • d rone’s body shifting and • drone’s body vibration • The system obtained high performance • at different distances, • in different environments, and • with different types of drones.

Matthan : Drone Presence Detection by y Id Identifying Physical Sig ignatures in in the Drone’s RF F Communicati tion Phuc Nguyen , Hoang Truong, Mahesh Ravindranathan, Anh Nguyen, Richard Han, Tam Vu