The high rate of intrusion of private and controlled air environments by unmanned aerial vehicles (UAVs) pose a serious security threat to radar, visual and acoustics systems that may be used to carry surveillance or intrude confined spaces and installations. To address the misuse of recreation drone technology resulting from the proliferation of privately owned UAVs, we investigate in this paper how to prevent UAV privacy intrusion, damage and security threats around confined spaces using a pre-trained Convolutional Neural Network (CNN) Image Recognition model. A binary image classification model is tested to distinguish the presence of drones’ images in an aerial environment from other flying and aerial objects. Our experimental test was based on a pre-trained Transfer learning model, Mobile Net-V2, implemented with a binary classification achieving an accuracy of 93% in detection drones. Performance evaluation also shows a 93%, 90% and 94% score for the F1-score, Recall and Precision, respectively.