A New Hybrid Adaptive Deep Learning-Based Framework for UAVs Faults and Attacks Detection
Fadhila Tlili, Samiha Ayed, Lamia Chaari Fourati
Abstract
A resilient and guaranteed Unmanned Aerial Vehicles (UAVs) security framework should be designed to be secure against different types of attacks and faults. Recent developments have seen a proliferation of methods for improving UAVs security. Although, many studies proposed different approaches using artificial intelligence to enhance their security. Unfortunately, no study yet worked on an examination of an hybrid framework on UAVs faults and attacks and different architectures. Hence, our article aims to provide a prior detection results by proposing an hybrid adaptive framework for faults and attacks detection for UAVs applied on centralized and decentralized architectures. Our framework is based on two entry flows for faults and attacks in order to learn high-level features automatically from data. We validated our framework using deep-learning architectures. Finally, the empirical results show that our framework reached over 85% and 96,7% in term of accuracy for UAVs faults and attacks, respectively.