Improving GNSS Reliability for Autonomous Vehicles by Detecting and Classifying Spoofing Attacks with Shallow Learning Models

Spoofing attacks on the Global Navigation Satellite System (GNSS) present significant threats to the security and reliability of autonomous vehicle systems, vital in civilian and military operations. Spoofing involves manipulating a vehicle's navigation system with false GNSS signals, potentially causing it to deviate from its path or crash. This study uses supervised Machine Learning (ML) models to detect GNSS spoofing attacks. These models were developed and tested using a dataset of authentic GNSS signals collected from diverse locations. Additionally, three types of GNSS spoofing attacks-simple, intermediate, and advanced-were simulated to mimic real-world scenarios using specialized equipment. The results of our experiments demonstrate the effectiveness of supervised ML in detecting and mitigating GNSS spoofing attacks. Testing accuracy rates of approximately 99.2% and 99.1% were achieved for the Decision Tree (DT) and K-Nearest Neighbor (KNN) algorithms, respectively. These findings underscore the importance of our approach in enhancing the accuracy and reliability of GNSS-based location services for autonomous vehicles. This research offers valuable insights for improving GNSS implementations and lays the groundwork for future advancements in spoofing detection methods.