Researchers have developed a framework to detect and classify Global Navigation Satellite System (GNSS) spoofing attacks on autonomous vehicles. The system uses small language models to analyze driving states from GNSS and other sensing sources, converting them into structured narratives for spoofing detection and classification. In experiments, the system achieved high accuracy and outperformed large language models in terms of computational efficiency and resource utilization. --- Why it matters: This matters because autonomous vehicles rely on reliable GNSS positioning, and detecting spoofing attacks is crucial to prevent potential safety risks. The framework's ability to detect attacks in real-time with low computational resources makes it suitable for deployment on resource-constrained vehicular computing platforms. Source: https://arxiv.org/abs/2608.17092