The commercial drone market has substantially grown over the past few years. While providing numerous advantages in various fields and applications, drones also provide ample opportunities for misuse by irresponsible hobbyists or malevolent actors. The increasing number of safety/security incidents in which drones are involved has motivated researchers to find new and ingenious ways to detect, locate and counter this type of vehicles. In this paper, we propose a new method to detect frequency hopping spread spectrum - Gaussian frequency-shift keying (FHSS-GFSK) drone communication signals, in a non-cooperative scenario, where no prior information about the signals of interest is available. The system is designed to detect and retrieve data bit sequences through a compressive sampling approach, which includes the extraction of the reduced spectral information and a soft detection algorithm. The performance of the proposed approach is assessed in terms of bit error rate and compared with that of a Viterbi detector and a neural network-based detector. The effectiveness of the method described in the paper highlights the fact that current UAV communications are not infallible and present real security issues.