Enhancing fiber security using a simple state of polarization analyzer and machine learning

The paper focuses on the security of fiber-optic cable infrastructures by detecting vibrations using an optical state of polarization analyzer. The developed system can detect various security breaches. The system only detects abnormal events without any event classification. The proposed system relies on an analyzer evaluating optical polarization differences caused by mechanical or acoustic vibrations analyzed by a machine-learning model for real-time anomaly detection. The main goal of experiments is to find the best combination of the normalization method and anomaly detector. The proposed system achieves an F1-score over 95.65%, which proves the solution's suitability for protecting fiber-optic infrastructures.