Impact of Dataset and Model Parameters on Machine Learning Performance for the Detection of GPS Spoofing Attacks on Unmanned Aerial Vehicles

GPS spoofing attacks are a severe threat to unmanned aerial vehicles. These manipulate the true state of vehicles, potentially misleading system without raising alarms. Several techniques, including machine learning, have been proposed detect these attacks. Most studies applied learning models identifying best hyperparameters, using feature selection and importance ensuring that used dataset is unbiased balanced. However, no current discussed impact model parameters characteristics on performance models; therefore, this paper fills gap by evaluating regularization parameters, size, correlated features, imbalanced datasets six most commonly known techniques. Classification Regression Decision Tree, Artificial Neural Network, Random Forest, Logistic Regression, Gaussian Naïve Bayes, Support Vector Machine. Thirteen features extracted from legitimate simulated attack signals perform investigation. The evaluation was performed in terms four metrics: accuracy, probability misdetection, false alarm, detection. results indicate adversely affect model’s performance. also show Tree classifier has an accuracy 99.99%, detection 99.98%, misdetection 0.2%, alarm 1.005%, after removing tuned balanced dataset. Forest can achieve 99.94%, 99.6%, 0.4%, 1.01% similar conditions.