Machine-Learning-Aided Localization-Based Attack Detection in Movable Antenna Systems

Physical-layer (PHY) attacks increasingly pose a threat to location-based services. To secure the localization mechanism against PHY attacks, we propose a novel framework based on localization and user-and-attacker detection, with the help of unsupervised machine learning (ML) algorithms and multiple signal classification (MUSIC) spectra. Our proposed framework consists of two stages: i) uplink localization and detection; and ii) downlink secure transmission. Noticeably, in the proposed framework, a reciprocal relationship between the localization mechanism and the user/attacker detection is developed, where the localization supports the detection and vice versa. This reciprocal relationship allows wireless systems to detect localization attacks and further localize the attacker. Through simulation, we show the efficacy of combining localization and detection in the uplink. We then demonstrate the benefit of employing both localization and movable-antenna arrays for secure downlink transmission.