Machine Learning for Intrusion Detection of Cyber-Attacks in Unmanned Aerial Vehicles
Open AccessThe introduction of Unmanned Aerial Vehicles (UAVs) has revolutionized civilian and military aviation operations. Their vast and advantageous applications induce high-value proposition. The global UAV market projects a revenue of $102 billion by 2030, with a compound annual growth rate of 19.6%. In fact, the According to the 2023 Presidential Budget, the Department of Defense planned on spending $2.6 billion in unmanned systems (McNabb, 2023). Despite their tangible benefits and value proposition, UAVs are vulnerable to significant security weaknesses that could impact human safety and national security. Evidently, there is a direct relationship between the demand for UAV systems and the incentive for threat actors to conduct malicious cyber activity. As such, it is pivotal to design and develop effective countermeasures to prevent unauthorized UAV intrusions. This research is concerned with UAV security vulnerabilities that disrupt GPS signals and proposes a Machine Learning approach to detect intrusions of cyber-attacks in UAVs. More specifically, it leverages supervised machine learning to effectively detect intrusion of cyber-attacks on the UAV Attack dataset (Whelan, et. al., 2020) via binary and multi-class classification, while simultaneously aiming to identify a classifier that outperforms prior approaches using standard classification metrics. The research methodology was founded on the data mining process comprised of data collection and understanding, data preparation, modeling, validation, and evaluation. Within this construct, 11 popular classification algorithms were modeled against the UAV Attack Dataset (Whelan, et. al., 2020) to address the research questions and hypotheses. The contributions and conclusions of this research codify that ML approaches are effective for classifying intrusion detection of cyber-attacks in UAVs with 80% precision and accuracy. This research additionally postulates the UAV Attack dataset as a useful dataset for analyzing UAV network environments. Furthermore, it validates that Tree-Based ML algorithms are the most effective for classification purposes when compared against the other classifiers used in this research. Lastly, it provides context into some of the possible factors that contributed to rejecting or accepting each research hypothesis.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.