Evaluating Machine Learning Models for Attack Detection in GPS Datasets

The Global Positioning System (GPS) has emerged as an omnipresent technology with applications spanning transportation, agriculture, disaster management, and various other domains. Nevertheless, the reliability of GPS signals is susceptible to being compromised by spoofing attacks, which have the potential to disseminate misleading location data. This study undertakes an evaluation and comparison of the efficacy of three distinct machine learning models, namely Logistic Regression, Random Forest, and Support Vector Machine (SVM), in the realm of identifying and thwarting spoofing attacks within GPS datasets. The process entails the utilization of a dataset encompassing both authentic and simulated GPS signals. This dataset is subjected to meticulous preprocessing, normalization, and subsequent division to facilitate the training and testing phases of the chosen models. The evaluation of these models encompasses a comprehensive array of performance metrics, encompassing accuracy, precision, recall, F1-score, and ROC AUC, to holistically gauge their effectiveness. The findings of this investigation unveil that, although both the Random Forest and SVM models yield heightened accuracy rates, the Logistic Regression model furnishes more conservative outcomes, potentially engendering enhanced generalizability when faced with uncharted data instances. This scholarly inquiry thus augments the comprehension of the utility of machine learning techniques in fortifying the integrity of the GPS infrastructure against the escalating specter of spoofing threats.