Drones are becoming an increasing part of the ever-connected society that we currently live in. Drones are used for delivering packages, geographic surveying, assessing the health of crops or just good old fashioned fun. Drones are excellent tools and their uses are expected to expand in the future. Yet, drones can be easily misused for malicious purposes if drone security isn't taken more seriously. One of the bigger problems drones have been causing lately is that they are being used to capture images or video of disasters, such as wildfires and in doing so get in the way of the relief effort. They also have caused several problems by flying too close to airports. These drones are usually too small for radar to pick up and are often discovered by visual means and by that time it is too late. One defense against this has been GPS designated no fly zones, however, this can be easily overcome by spoofing the GPS signal to make the drone think it is in a safe area to fly.

In this paper, I examine ways of detecting the presence of a drone using machine learning models by recording the RF spectrum during a drone’s flight and then feeding the raw data into a machine learning model. This could be used around airports or even on the airplanes themselves to detect the presence and/or approach of a drone. Specifically, I examine two very popular consumer drones and their transmitters: The 3D Robotics Solo and the DJI Phantom 2. These two types of drones are unique in the way that they send and receive signals to the transmitter. I show that machine learning models, once trained, can detect drone activity in the RF spectrum. However, more work is needed in order to improve the detection rate of these models so that they may be employed in a practical manner.