This PhD thesis focuses on capturing the behavioral data of different pilots while flying UAVs and using machine learning algorithms to identify the pilot. The objective is to develop a mechanism that enables UAVs to recognize the pilot in order to avoid being hijacked by non-authorized pilots. The thesis includes the testing and classification of 20 features against 22 classifiers, with Random Forest being selected as the best classifier. The results show that a single feature can score an accuracy of 95.9% in pilot identification, while the overall performance of combined features is 99.40%. The thesis also explores the reduction of random forest trees to save power and memory.
![[object Object]](/_next/image/?url=%2F_next%2Fstatic%2Fmedia%2Flogo.6d15ce61.png&w=640&q=75){kind=small}
![[object Object]](/_next/static/media/star-bottom.7253800d.svg)
