Processing audio signals recorded from a microphone array embedded in an unmanned aerial vehicle (UAV) has received increasing research interest in the recent years and has been referred to as drone audition. An important field of application is search and rescue, where humans in disaster areas need to be quickly found. UAVs equipped with high-resolution cameras have already been used in humanitarian responses, while audio-based UAV-embedded sound localisation remains an open research challenge. Microphones could provide a critical complementary modality to vision in situations where visual feedbacks are limited due to bad lighting conditions (night, fog) or obstacles limiting the field of view. This paper provides an overview of the technical and methodological challenges faced by drone audition in the context of search and rescue and presents two publicly available datasets that aim at fostering research in this area. Some localisation and noise-reduction results obtained using baseline methods are also presented. While static localisation of speech sources from a distance of four meters can be efficiently achieved, in-flight localisation from larger distances remains a challenge.