International Journal of Applied Information Systems |
Foundation of Computer Science (FCS), NY, USA |
Volume 12 - Number 47 |
Year of Publication: 2025 |
Authors: Oluwaseun Adesola-Zion , Kazeem B. Adedeji |
![]() |
Oluwaseun Adesola-Zion , Kazeem B. Adedeji . Development of a Distributed Denial of Service Attack Detection Scheme for Multi-UAV Network. International Journal of Applied Information Systems. 12, 47 ( Mar 2025), 16-24. DOI=10.5120/ijais2025452017
Unmanned Aerial Vehicle (UAV) networks are susceptible to several cyber attack due to the broadcast nature of the wireless communication architecture between the UAV and the ground station. Among these threats, Distributed Denial-of-Service (DDoS) attacks pose significant risks to UAV networks. This study develops an effective machine learning-based scheme for detecting DDoS attacks in UAV networks. A comprehensive, labeled network traffic dataset, encompassing both normal and malicious traffic, was curated and preprocessed through normalization and the removal of missing values. Three ensemble classifiers were developed for attack detection. Classifier 1 combines Logistic Regression (LR) and Decision Tree (DT), Classifier 2 integrates Random Forest (RF) and DT and Classifier 3 leverages a hybrid of LR, DT, and RF. The classifiers were trained and evaluated using a dataset split into 70% training, 10% validation, and 20% test subsets. Feature extraction technique was employed to identify key characteristics of network traffic essential for detecting attack patterns. The classifiers' performance was assessed using metrics such as accuracy, precision, recall, F1-score, ROC curve, loss function, and epoch analysis. Results showed that Classifier 2 achieved the best performance, with 97.05% accuracy, 98.79% precision, and a 97.27% F1-score, demonstrating its robustness in detecting DDoS attacks. Classifier 3 exhibited comparable performance, with 97.09% accuracy, 97.41% precision, and 97.34% F1-score, but a slightly higher loss value, making it slightly less robust. Classifier 1, while achieving reasonable accuracy (87.68%) and precision (97.99%), showed weaker recall (79.17%) and F1-score (87.58%), indicating limited reliability. This study has shown that the detection accuracy of DDoS attack in UAV networks can be improved with the use of ensemble-based methodology.