A Machine Learning Approach for Intrusion Detection in Drone Communication Networks
Book Chapter
Saripalli, J, Gangwani, P, Upadhyay, H et al. (2027). A Machine Learning Approach for Intrusion Detection in Drone Communication Networks
. 685 LNICST 96-109. 10.1007/978-3-032-22542-9_7
Saripalli, J, Gangwani, P, Upadhyay, H et al. (2027). A Machine Learning Approach for Intrusion Detection in Drone Communication Networks
. 685 LNICST 96-109. 10.1007/978-3-032-22542-9_7
From logistics to defense, UAVs are evolving industries, however their communication networks remain vulnerable to cyber-attacks. This work presents a comparative assessment of five machine learning-based supervised algorithms for the detection of intrusion in drone communication networks. Considering real UAV network traffic data from the ECU-IoFT dataset, we performed intrusion detection by employing the following machine learning and deep learning algorithms, Support Vector Machine, Decision Tree, Random Forest, XGBoost, and Multi-Layer Perceptron, for a binary classification task of WPA2-PSK WIFI Cracking Attack (referred to as “Attack” in this paper) versus Normal traffic. Feature engineering was used in the methodology to avoid data leakage, so that models learn only from legitimate packet-level characteristics. Evaluating using 10-fold stratified cross-validation showed that the Decision Tree achieved the highest accuracy, at 95.14%, while Random Forest and XGBoost performed comparably at 95.13% and 95.11%, respectively. All tree-based methods demonstrated balanced precision and recall in both traffic classes, while the recall in detecting Attack was as high as 90%. These results demonstrate that the employment of tree-based classifiers allows for effective threat detection in real time on resource-constrained UAV platforms.
