Global Navigation Satellite Systems (GNSS) continue to be a predominant technology for the enablement of tactical maneuver in military operations, yet remain vulnerable to electronic warfare (EW) threats, particularly jamming and spoofing. This research develops and evaluates a machine learning (ML)-based GNSS anomaly detection framework, integrating Long Short-Term Memory (LSTM) autoencoders into an adaptive pipeline capable of identifying signal anomalies associated with EW attacks. The solution leverages real-world GNSS data from Estonia's Positioning System Network (ESTPOS), incorporating statistical filtering and per-satellite feature analysis to improve detection fidelity. Evaluation against unlabeled but geopolitically verified jamming intervals demonstrates timely and reliable detection of diverse jamming events, outperforming static threshold-based methods in operational conditions. Our solutions modular design ensures suitability for embedded military deployments and future extensibility, including potential integration of spoofing detection and multi-sensor data fusion.