Smart Self-Diagnosis Method for GPS Attacks and Safety Faults in UAVs

Unmanned aerial vehicles (UAVs) have been used in a variety of applications in safety-critical domains such as logistics, transportation, and defense. However, safety assurance is needed for the widespread development and operation of UAVs in urban areas where a failure may lead to physical harm to people, environment, or property. Since UAV s are highly interconnected, security is an important concern for the acceptance of safety-critical UAV applications. Thus, it is needed to ensure that UAV s are protected from attacks, e.g., Global Position System (GPS) signal spoofing, and unauthorized manipulation of flight controls, caused by malicious external agents aiming to compromise system/data confidentiality, integrity, or availability. Artificial Intelligence techniques such as Machine Learning classifiers can be used to support the identification of UAV s attacks. This study presents the development of safety and security diagnostic methods for UAVs. Two multiclass approaches are proposed, focusing on identifying safety-related failures and security threats such as GPS spoofing and jamming attacks. Various machine learning algorithms are employed, and evaluation metrics demonstrate high accuracy rates across models. The study underscores the importance of addressing both safety and security aspects in UAV s operations and provides pre-processing code for reproducibility. The developed methods contribute to enhancing UAV resilience and understanding safety and security measures in unmanned aerial operations.