Contents I. Introduction
The health status of electric cables is very important for national construction, which is infected by external third-party interference seriously. Compared with traditional scheme of Manned patrol, DAS based on the phase-sensitive optical time-domain reflectometry (-OTDR) technology has advantages including high sensitivity and precision, dynamic detection, all-weather monitoring, long-distance monitoring, and cost-efficient[1–3], which is very popular for intrusion detection in the third-party damage warning for perimeter[4], large structure healthy monitoring[5], and oil pipelines[6]. So it is very suitable for third-party interference monitoring of communication cables[7]. Deep learning method can relieve the false alarm rate (FAR) problem caused by background noise along the fiber link. But machine learning methods based on the feature extraction rely heavily on expert knowledge and have poor recognition rate[8, 9]. Other deep learning method such as 2-D CNN[10, 11],LSTM[12], are unable to meet the requirement of real-time monitoring.
