GenAI-Based Jamming and Spoofing Attacks on UAVs

Recently, aerial vehicles have been more connected than ever, where there are many types of the vehicles. Unmanned aerial vehicles (UAVs) operate on various environments with different technologies that are subject of many attacks. Creating effective intrusion detection systems against such attacks have been a significant challenge since there is a lack of sufficient attack data that can be used to design an intrusion detection system with advanced computing algorithms. In this research, we propose a novel framework to create attacks data for UAVs by using generative artificial intelligence algorithms. We use Variational Autoencoder, Gaussian Copula, Denoising Diffusion Probabilistic Model (DDPM), and Conditional Tabular Generative Adversarial Network to create synthetic attack data. Specifically, jamming and spoofing attacks on UAVs are generated to fool intrusion detection systems that may be implemented on UAVs. Experimental evaluations show that synthetically generated attack data reduces the accuracy of intrusion detections if the system was trained with inadequate attack data. Additionally, analysis results show that DDPM emerged as the most effective model for generating attack data, leading to F1 score reductions of 21% for jamming and 28% for spoofing attacks. This research highlights the need for more robust and adaptive intrusion detection systems that can be created with synthetic data. Thus, sustainable computing systems on UAVs will be achieved.